--- Log opened Mon Aug 05 00:00:01 2013 | ||
shogun-buildbot | build #1653 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1653 | 00:03 |
---|---|---|
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has joined #shogun | 00:05 | |
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/9840053 | 00:05 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has left #shogun [] | 00:05 | |
shogun-buildbot | build #838 of rpm1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/838 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 00:06 |
shogun-buildbot | build #1156 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1156 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 00:15 |
-!- travis-ci [~travis-ci@ec2-54-225-54-153.compute-1.amazonaws.com] has joined #shogun | 00:30 | |
travis-ci | [travis-ci] it's Sergey Lisitsyn'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/9840702 | 00:30 |
-!- travis-ci [~travis-ci@ec2-54-225-54-153.compute-1.amazonaws.com] has left #shogun [] | 00:30 | |
shogun-buildbot | build #1463 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1463 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 01:58 |
-!- zxtx [~zv@24-221-57-180.pools.static.spcsdns.net] has joined #shogun | 02:54 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 02:55 | |
shogun-buildbot | build #424 of nightly_none is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_none/builds/424 | 03:01 |
-!- zxtx [~zv@24-221-57-180.pools.static.spcsdns.net] has quit [Ping timeout: 256 seconds] | 03:05 | |
shogun-buildbot | build #479 of nightly_default is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/479 | 03:39 |
shogun-buildbot | build #414 of nightly_all is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_all/builds/414 | 05:19 |
-!- lisitsyn [~lisitsyn@5-28-21-18.clients.tlt.100megabit.ru] has quit [Quit: Leaving.] | 07:24 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has joined #shogun | 08:18 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 08:18 | |
@iglesiasg | good morning | 08:24 |
@iglesiasg | HeikoS: hey there! | 08:26 |
@iglesiasg | HeikoS: The thing I mentioned with GaussianDistribution unit test may be happening in travis too | 08:26 |
@iglesiasg | HeikoS: https://travis-ci.org/shogun-toolbox/shogun/builds have a look at the build history | 08:26 |
@iglesiasg | HeikoS: From build 1626 onwards build for libshogun with unit test is always grey | 08:27 |
@iglesiasg | the one that is immediately before finished with green though, https://travis-ci.org/shogun-toolbox/shogun/builds/9825642 | 08:27 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 08:46 | |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 8570ed0 / src/shogun/machine/gp/ (2 files): https://github.com/shogun-toolbox/shogun/commit/8570ed076e78287acd3fb609c1d58e7a3f86bec3 | 08:46 |
shogun-notifier- | shogun: method actually uses CGaussianDistribution which needs Eigen3 so ifdef with that | 08:46 |
shogun-notifier- | shogun: van51 :develop * 313bba5 / / (8 files): https://github.com/shogun-toolbox/shogun/commit/313bba5b3e17e2f313cc54e1f7999b2ae35bedee | 08:52 |
shogun-notifier- | shogun: Normalization in CHashedDocConverter class | 08:52 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * d5f6740 / / (8 files): https://github.com/shogun-toolbox/shogun/commit/d5f6740668474fb4684758993e488fb3d2bbe0a9 | 08:52 |
shogun-notifier- | shogun: Merge pull request #1330 from van51/develop | 08:52 |
shogun-notifier- | shogun: | 08:52 |
shogun-notifier- | shogun: Normalization in CHashedDocConverter class | 08:52 |
-!- foulwall [~user@110.16.165.220] has joined #shogun | 08:59 | |
shogun-buildbot | build #1655 of deb1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1655 blamelist: van51 <vangelis_51@hotmail.com> | 09:01 |
shogun-buildbot | build #1656 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/1656 | 09:04 |
shogun-buildbot | build #1157 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1157 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:05 |
shogun-buildbot | build #839 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/839 | 09:13 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has joined #shogun | 09:21 | |
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/9848933 | 09:21 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has left #shogun [] | 09:21 | |
-!- sonne|work [~sonnenbu@91-64-72-127-dynip.superkabel.de] has joined #shogun | 09:30 | |
shogun-buildbot | build #1464 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1464 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:34 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has joined #shogun | 09:41 | |
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/9849023 | 09:41 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has left #shogun [] | 09:41 | |
sonne|work | foulwall: ping ping ping | 09:41 |
shogun-buildbot | build #1158 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1158 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:43 |
foulwall | ping sonne|work | 09:44 |
sonne|work | foulwall: hey | 09:45 |
sonne|work | foulwall: we need to make the dimred examples interactive | 09:46 |
foulwall | now working for tree https://www.biomedcentral.com/content/figures/1471-2105-5-126-1-l.jpg | 09:46 |
sonne|work | foulwall: please talk to lisitsyn abou tit | 09:46 |
sonne|work | foulwall: nice | 09:46 |
sonne|work | foulwall: we have hierarchical clustering in shogun and an example so we can then display that | 09:47 |
foulwall | ok sonne|work, I'll show you the tree soon, and add some backend code to it. | 09:47 |
sonne|work | foulwall: I guess I should then try to add some algorithms and data sets and then we will see what is missing | 09:56 |
sonne|work | (for the toy examples) | 09:56 |
sonne|work | foulwall: the only other stuff you could do is add some way of running asp / arts also from the web | 09:57 |
sonne|work | they both need some string to be uploaded | 09:57 |
sonne|work | and some curve to be drawn | 09:57 |
sonne|work | not too difficult | 09:57 |
shogun-buildbot | build #1465 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1465 blamelist: Soeren Sonnenburg <sonne@debian.org> | 09:58 |
sonne|work | iglesiasg: ok now? | 09:58 |
@iglesiasg | sonne|work: what? | 09:59 |
sonne|work | the get_distance_matrix? | 09:59 |
@iglesiasg | sonne|work: I didn't try it again, did you change something? | 10:00 |
sonne|work | man | 10:00 |
sonne|work | I even pinged you | 10:00 |
sonne|work | so yes sure try it and my question remains | 10:00 |
@iglesiasg | sonne|work: I saw you said yesterday about it was a pthread problem | 10:00 |
sonne|work | do you have a good name for a common base class of CKernel & CDistance | 10:00 |
@iglesiasg | similarity measure | 10:01 |
sonne|work | these things are somewhat similar so I would try to refactor them | 10:01 |
sonne|work | iglesiasg: hmmh distance is rather dissimilarity and kernel similarity | 10:01 |
@iglesiasg | I guess that is a matter of taste :) | 10:02 |
@iglesiasg | I mean why increasing magnitude must mean what you are trying to describe | 10:03 |
@iglesiasg | decreasing magnitude is equally describing | 10:03 |
@iglesiasg | and distance=0 means they are as similar as they can get | 10:03 |
@iglesiasg | anyway, I am not picky with names (to some extent) | 10:04 |
@iglesiasg | sonne|work: why did you add init_shogun_with defaults and exit_shogun in the unit test btw? | 10:05 |
@iglesiasg | https://github.com/shogun-toolbox/shogun/commit/d474227dac6e804fd8376877237f5b6bf49af67a | 10:05 |
@iglesiasg | sonney2k: and yep, it is ok now :) | 10:05 |
sonne|work | iglesiasg: I wanted to do display_matrix() | 10:05 |
@iglesiasg | sonney2k: it doesn't work otherwise? | 10:06 |
@iglesiasg | sonne|work: | 10:06 |
sonne|work | ermm no idea | 10:06 |
foulwall | gotcha sonne|work | 10:06 |
@iglesiasg | let me try | 10:07 |
sonne|work | foulwall: but for the asp/arts applications you have to be *very* aware about security issues like upload a too big file (-> crash) etc | 10:08 |
sonne|work | iglesiasg: ok so CSimilarity then? | 10:08 |
@iglesiasg | sonne|work: it is fine for me | 10:09 |
@iglesiasg | sonney2k: and display matrix works fine with init_shogun yeah | 10:09 |
sonne|work | iglesiasg: with or without? | 10:10 |
@iglesiasg | I keep on speaking to your home alter ego | 10:10 |
@iglesiasg | sonne|work: without, sorry | 10:10 |
@iglesiasg | sonne|work: so the holidays have ended now? | 10:10 |
sonne|work | yea :( | 10:11 |
@iglesiasg | I am sorry :( | 10:11 |
sonne|work | in particular since it is supposed to be 29 C today in Berlin | 10:11 |
sonne|work | and 33 C Tue/Wed | 10:11 |
@iglesiasg | awesome | 10:11 |
@iglesiasg | I think it is a really cool city, I was even surprised with it! | 10:12 |
sonne|work | iglesiasg: which reminds we that we still don't have our video recordings online | 10:13 |
sonne|work | wiking: any progress? | 10:13 |
@iglesiasg | yeah! I have to see HeikoS video again and take notes :D | 10:14 |
* sonne|work too | 10:15 | |
@wiking | sonne|work: with? | 10:18 |
sonne|work | the video cutting and our youtube channel :) | 10:18 |
@wiking | yeah i've cutted the morning session | 10:20 |
@wiking | the shogun channel should be registered on a common shogun email | 10:25 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 10:28 | |
lisitsyn | haha sonne|work now you too | 10:28 |
lisitsyn | *work* | 10:28 |
lisitsyn | *evil haha* | 10:28 |
sonne|work | lisitsyn: baeh! | 10:29 |
lisitsyn | iglesiasg: is it happening right now? | 10:29 |
sonne|work | wiking: what email do we want for that? | 10:29 |
sonne|work | wiking: cool! | 10:29 |
sonne|work | sth like shogun@shogun-toolbox.org ? | 10:30 |
sonne|work | or what? | 10:30 |
@wiking | yeah | 10:31 |
@wiking | something like that | 10:31 |
* wiking is debugging octave on ubuntu 12.04 | 10:32 | |
sonne|work | wiking: and us 5 being the ones who receive emails on that address? | 10:32 |
@wiking | yeah | 10:32 |
-!- hushell [~hushell@c-24-21-169-136.hsd1.or.comcast.net] has joined #shogun | 10:32 | |
sonne|work | ok | 10:33 |
sonne|work | wiking: I will create that then tonight | 10:33 |
@wiking | cool | 10:33 |
@wiking | then i'll create the shogun youtube channel | 10:33 |
hushell | iglesiasg: Hi, I sent a PR, could you help me check the MAPInference.h? Do I make things right for the opaque pointer? | 10:35 |
lisitsyn | iglesiasg: you are the opaque pointer guy now ;) | 10:35 |
hushell | haha | 10:35 |
@wiking | lisitsyn: do u have still octave | 10:35 |
@wiking | on your machine..? | 10:35 |
lisitsyn | wiking: I can run it on a virtual machine with debian | 10:36 |
lisitsyn | just let me know what to test | 10:36 |
sonne|work | lisitsyn: if we are into gsoc next year opaque pointering whole shogun would be some task | 10:36 |
@wiking | lisitsyn: ah no i was hoping u have the same machine that u had the last day where octave was detected fine | 10:37 |
sonne|work | wiking: btw can you somehow extract your cmake history and merge it into develop? | 10:37 |
@iglesiasg | hushell: hey! I will read it later this evening, ok? | 10:38 |
@wiking | sonne|work: i'll rebase it first with develop | 10:38 |
sonne|work | wiking: I would very much prefer that you do it there and I could already start switching buildbots | 10:38 |
@wiking | and then i'll merge or something | 10:38 |
sonne|work | wiking: just no mess please (I thought you are just working on a single file...) | 10:38 |
hushell | iglesiasg: sure. I am going to sleep right now. Take your time :) | 10:38 |
@wiking | sonne|work: no it's serveral files | 10:39 |
@wiking | but it still needs work | 10:39 |
@wiking | so i'll do some more tests | 10:39 |
@wiking | before anything | 10:39 |
sonne|work | wiking: but for example libshogun will certainly work | 10:39 |
@wiking | yeah for sure | 10:40 |
sonne|work | wiking: and cygwin is failing with configure with 30% chance | 10:40 |
sonne|work | so I would prefer to switch ASAP there | 10:40 |
lisitsyn | wiking: I can test this thing later tonight then if it is still relevant by that time | 10:40 |
sonne|work | lisitsyn: please talk to foulwall about the dimred demos! | 10:41 |
@wiking | sonne|work: can we install VS on that cygwin bot? | 10:42 |
sonne|work | wiking: I guess so | 10:42 |
@wiking | sonne|work: we could try to build shogun with VS :P | 10:43 |
sonne|work | wiking: btw how difficult is it with travis to add another python modular build with python3.x? | 10:43 |
sonne|work | wiking: it will kaboom for sure | 10:43 |
sonne|work | wiking: but yes let me write an email and ask | 10:43 |
@wiking | sonne|work: i guess 12.04 has no python3 or? | 10:43 |
@wiking | it does | 10:43 |
@wiking | in that case it's not complicated at all | 10:43 |
sonne|work | wiking: no idea - lisitsyn do you remember? | 10:45 |
lisitsyn | wiking: you can install python3 indeed | 10:48 |
lisitsyn | default is not 3 though | 10:48 |
sonne|work | wiking: well then :D | 10:50 |
sonne|work | wiking: would be nice to have - gsomix did invest so much time in this - lets not have code rot kill that | 10:51 |
-!- van51 [~van51@ppp-94-66-23-53.home.otenet.gr] has joined #shogun | 10:56 | |
van51 | hello | 10:57 |
sonne|work | hey van51! | 11:02 |
sonne|work | how is it going | 11:03 |
sonne|work | any exciting benchmark news? | 11:03 |
foulwall | ping lisitsyn | 11:05 |
lisitsyn | foulwall: pong | 11:05 |
foulwall | hey:) | 11:06 |
lisitsyn | hey | 11:06 |
lisitsyn | I guess you want to talk about something interactive with DR | 11:06 |
foulwall | yeah, how about add some more data set? | 11:07 |
foulwall | user upload is not practical for security reasons. | 11:07 |
foulwall | lisitsyn: any ideas? | 11:08 |
lisitsyn | foulwall: yes you can change algorithm parameters | 11:09 |
lisitsyn | for example | 11:09 |
lisitsyn | for LLE and LTSA | 11:09 |
lisitsyn | you can change k | 11:09 |
lisitsyn | foulwall: reasonable values range of k is [6, number of vectors] | 11:09 |
lisitsyn | that's the number of neighbors | 11:10 |
-!- thoralf [~thoralf@enki.zib.de] has joined #shogun | 11:10 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 11:10 | |
foulwall | lisitsyn: Ok, :)I'll do that. | 11:10 |
thoralf | Hey. | 11:11 |
lisitsyn | foulwall: how do you compute embeddings now? | 11:13 |
van51 | sonney2k: hey! | 11:14 |
lisitsyn | you just load it from json data I had, right? | 11:14 |
van51 | sonney2k: well on the webspam, on about ~10k examples, the converter is much faster | 11:14 |
foulwall | lisitsyn: yes, | 11:14 |
van51 | sonney2k: it takes like the 1/3 ~ 1/2 of the time | 11:14 |
van51 | sonne|work: ^ | 11:15 |
van51 | sonne|work: but I had trouble running it on more examples, as it was running out of memory | 11:15 |
sonne|work | van51: including conversion time or excluding? | 11:17 |
van51 | sonney2k: including | 11:17 |
sonne|work | but wrt out of memory I would have guessed so | 11:17 |
sonne|work | van51: and excluding? | 11:17 |
-!- thoralfk [~thoralf@enki.zib.de] has joined #shogun | 11:17 | |
sonne|work | van51: I guess conversion time is just a few secs though? | 11:18 |
van51 | sonne|work: well, the training time of the SparseFeatures returned is instantaneous | 11:18 |
-!- thoralf [~thoralf@enki.zib.de] has quit [Remote host closed the connection] | 11:18 | |
van51 | sonne|work: and the conversion doesn't take too much | 11:18 |
sonne|work | van51: ok then this would make a nice graph/blog post | 11:19 |
sonne|work | van51: btw did you manage to get the string kernel based variant running? | 11:20 |
van51 | sonne|work: I had attempted it when you made that change, but it was crashing again | 11:21 |
van51 | sonne|work: and I didn't have the chance to further test it, had some busy days :S | 11:21 |
van51 | sonne|work: I can try it again now | 11:21 |
van51 | sonne|work: what I worked on a bit is the language detection | 11:22 |
-!- hoijui [~hoijui@adsl-84-227-37-53.adslplus.ch] has joined #shogun | 11:22 | |
sonne|work | van51: if it is still crashing (please run it with valgrind on some really small data) then I guess I will have to debug it - best would be you could craft some minimal unit test with a failing example then | 11:23 |
sonne|work | so I can concentrate on a fix | 11:23 |
sonne|work | van51: (excited) how is it going with the language detection? | 11:23 |
@iglesiasg | did kernel_anova_modular.py start failing recently? | 11:24 |
sonne|work | that is also a nice application and demo that can easily be integrated with foulwalls asp/arts based applications on the web :) | 11:24 |
sonne|work | iglesiasg: errm | 11:24 |
@iglesiasg | https://travis-ci.org/shogun-toolbox/shogun/jobs/9849933 | 11:24 |
sonne|work | I touched it | 11:24 |
sonne|work | but I was 'just' using gsomix' csv reader | 11:24 |
sonne|work | thought that it was actaully working | 11:24 |
van51 | sonne|work: very well actully. with just a small subset it achieves like 99% accuracy | 11:25 |
sonne|work | van51: yeah I guessed so :) | 11:25 |
sonne|work | van51: which n-gram length? | 11:26 |
van51 | sonne|work: 8 | 11:26 |
sonne|work | hmmhh pretty long! | 11:26 |
sonne|work | van51: btw I was thinking that maybe having ngrams of size 1...8 might actually also be useful ( I mean mixtures) | 11:27 |
sonne|work | van51: maybe you try with CombinedDotFeatures to have 2 of your HashedDocFeatures | 11:27 |
sonne|work | one with say n=4 and one wiht n=8 | 11:27 |
sonne|work | but if it is that good already | 11:28 |
sonne|work | it is probably sufficient | 11:28 |
van51 | sonne|work: it is good on those "large documents" | 11:28 |
sonne|work | ahh ok | 11:28 |
van51 | sonne|work: it had some trouble in small sentences | 11:28 |
sonne|work | we should try short sentences | 11:28 |
van51 | sonne|work: when I was testing it | 11:28 |
van51 | sonne|work: maybe with the reduced size it will be better there as well | 11:28 |
sonne|work | so basically you extract from the wikipedia documents shorter substrings (maybe even full sentences) | 11:29 |
van51 | sonne|work: I was thinking of testing that with the quadratic hasheddoc features once they are merged | 11:29 |
sonne|work | or something like x words | 11:29 |
sonne|work | and then train / test on that | 11:29 |
van51 | sonne|work: hmm ok can do :D | 11:29 |
-!- hoijui [~hoijui@adsl-84-227-37-53.adslplus.ch] has quit [Quit: Leaving] | 11:31 | |
van51 | sonne|work: I would like to port the classes to the modular interfaces first | 11:34 |
sonne|work | van51: ohh that port takes 5 minutes | 11:35 |
van51 | sonne|work: I just add the appropriate includes and rename it? | 11:35 |
sonne|work | van51: lets just do it now | 11:35 |
sonne|work | yes exactly | 11:35 |
sonne|work | search for CombinedDotFeatures in interfaces/modular/Feature* | 11:36 |
sonne|work | then copy and change your code | 11:36 |
sonne|work | err change to your class names | 11:36 |
sonne|work | recompile and try | 11:36 |
sonne|work | and PR :) | 11:36 |
sonne|work | van51: just do that now and later please explain what https://github.com/shogun-toolbox/shogun/pull/1342 is | 11:37 |
gsomix | good morning? =___=' | 11:37 |
sonne|work | gsomix: hey there! | 11:42 |
gsomix | sonne|work, sth wrong with csv reader? | 11:42 |
sonne|work | gsomix: not clear yet | 11:42 |
sonne|work | gsomix: maybe I am just using it wrong | 11:42 |
gsomix | sonne|work, I'm compiling python_modular now, then I can check. | 11:43 |
sonne|work | gsomix: I guess it might be my C order change | 11:43 |
sonne|work | not sure though | 11:44 |
van51 | sonne|work: the idea behind that PR is to combine tokens in a specific window of size n+s, where n is the number of the tokens to combine and s is how many tokens you 're allowed to skip | 11:44 |
sonne|work | I would say it makes much more sense to have ascii data in a way | 11:44 |
sonne|work | that each example is in one line | 11:44 |
sonne|work | so whatever other tools might do we shoudl stick to that | 11:45 |
van51 | sonne|work: I have an inline note there that explain this notion of skipping | 11:45 |
sonne|work | van51: yeah I was about to ask | 11:45 |
van51 | sonne|work: also all combination of smaller size are addeed | 11:46 |
van51 | combinations*, added* | 11:46 |
sonne|work | van51: btw I was thinking that maybe it is better to have a *fixed* seed for the hashing - did olivier/benoit say sth about it? | 11:47 |
van51 | sonne|work: no that hasn't been discussed | 11:48 |
van51 | sonne|work: but it is fixed atm | 11:48 |
sonne|work | van51: ahh ok then | 11:48 |
@wiking | sonne|work: ok i was away but yeah i'll add python3 modular test on travis | 11:48 |
sonne|work | van51: what you are doing is called (in its extreme) the subsequence kernel | 11:48 |
sonne|work | van51: very useful | 11:48 |
sonne|work | van51: the skipping and combination I mean | 11:49 |
sonne|work | van51: if you allow n=3 it would take anything from n=1 up to 3 right? | 11:49 |
shogun-notifier- | shogun: Fernando Iglesias :develop * d1b8776 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/d1b877642b86a06ba3bea73a30e5bc5e47aaa9f4 | 11:49 |
shogun-notifier- | shogun: Drop deprecated dense computation of all the outer products in LMNN and its unit test | 11:49 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 50edb28 / src/shogun/metric/LMNNImpl.h: https://github.com/shogun-toolbox/shogun/commit/50edb28485b30d9dfad4ec7da423cc809fec9a34 | 11:49 |
shogun-notifier- | shogun: Modify ImpostorNode in LMNNImpl from class to struct | 11:49 |
shogun-notifier- | shogun: Fernando Iglesias :develop * a9137d5 / src/shogun/metric/LMNN.cpp: https://github.com/shogun-toolbox/shogun/commit/a9137d50c3d9ca3c23544c31aa0cd9005d6e1c6e | 11:49 |
shogun-notifier- | shogun: Better debug output in LMNN | 11:49 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 5757928 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/57579284463d1187984b5ed1c2ae9e6f6f66dfb3 | 11:49 |
shogun-notifier- | shogun: Merge pull request #1358 from iglesias/feature/lmnn | 11:49 |
shogun-notifier- | shogun: | 11:49 |
shogun-notifier- | shogun: Feature/lmnn | 11:49 |
sonne|work | van51: https://cs.fit.edu/~pkc/apweb/related/lodhi-jmlr02.pdf | 11:50 |
-!- van511 [~van51@ppp-94-66-42-68.home.otenet.gr] has joined #shogun | 11:51 | |
van511 | sonne|work: yeah it would create anything from n=1..3 | 11:52 |
-!- van51 [~van51@ppp-94-66-23-53.home.otenet.gr] has quit [Ping timeout: 256 seconds] | 11:53 | |
sonne|work | van511: that is the subsequence kernel paper https://cs.fit.edu/~pkc/apweb/related/lodhi-jmlr02.pdf | 11:53 |
sonne|work | just if you are interested | 11:53 |
sonne|work | IMHO their approach is not so useful | 11:54 |
sonne|work | yours sounds *MUCH* more reasonable | 11:54 |
van511 | sonne|work: hehe it's vw's approach for ngrams as well | 11:55 |
van511 | sonne|work: at least based on some comments, because their code is not that readable | 11:55 |
shogun-buildbot | build #841 of rpm1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/841 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 11:56 |
sonne|work | van511: the approach you are doing you mean? | 11:56 |
van511 | sonne|work: yes | 11:56 |
sonne|work | yeah ok | 11:57 |
sonne|work | I guess that would also work very nicely with short sentences | 11:57 |
sonne|work | van511: btw is this to be merged or are you still testing it? | 11:57 |
gsomix | sonney2k, yes, problem with order. | 11:57 |
gsomix | sonne|work, ^ | 11:58 |
van511 | sonne|work: not yet, I have to remove a previous commit and clean some things | 11:58 |
van511 | sonne|work: it will be ready in a while though | 11:58 |
sonne|work | gsomix: so with fortran order it would work right? | 11:59 |
gsomix | sonne|work, yep. | 11:59 |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 11:59 | |
-!- van511 is now known as van51 | 12:00 | |
gsomix | sonne|work, I'll make changes. and let Travis test it. | 12:01 |
sonne|work | gsomix: err wait | 12:01 |
sonne|work | I would rather transpose the ascii data | 12:01 |
sonne|work | I mean read it in write it transposed | 12:02 |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * e6c27f7 / .travis.yml,CMakeLists.txt,cmake/FindOctave.cmake,cmake/FindApacheAnt.cmake: https://github.com/shogun-toolbox/shogun/commit/e6c27f77bc9bccc4f8e0d13b0ac7ba0a880ece2a | 12:03 |
shogun-notifier- | shogun: Fix FindOctave for ubuntu 12.04 | 12:03 |
shogun-notifier- | shogun: add python3 modular test to travis | 12:03 |
gsomix | sonne|work, didn't get it. for what? | 12:03 |
@wiking | let see what travis does with this | 12:03 |
sonne|work | wiking: nice! | 12:03 |
sonne|work | gsomix: well I want to keep it C-order | 12:04 |
shogun-buildbot | build #842 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/842 | 12:04 |
sonne|work | gsomix: it just makes more sense for ascii | 12:04 |
sonne|work | gsomix: so if I would flip the data in the example file it will work | 12:04 |
@wiking | sonne|work: imo we should enable ccache for compiler and swig on buildbots | 12:04 |
sonne|work | wiking: it is enabled | 12:05 |
sonne|work | wiking: maybe not on yours :D | 12:05 |
shogun-buildbot | build #1159 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1159 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 12:05 |
@wiking | sonne|work: for swig as well? | 12:05 |
sonne|work | no | 12:05 |
@wiking | sonne|work: ccache-swig | 12:05 |
@wiking | sonne|work: it's coming with swig package | 12:05 |
sonne|work | I know | 12:05 |
sonne|work | but not enabled | 12:05 |
shogun-buildbot | build #1160 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1160 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 12:06 |
@wiking | ok i'll add an ENABLE_CCACHE to the cmake | 12:06 |
@wiking | which will be by default disabled | 12:06 |
@wiking | i still have some build problems on travis with ruby_modular | 12:06 |
@wiking | but first i'll try now to rebase the branch to develop | 12:07 |
sonne|work | wiking: no enable by default | 12:08 |
@wiking | ok the only problem with rebase actually is with | 12:08 |
gsomix | sonney2k, hm, ok. but I don't think that its good idea. SGMatrix stores data in fortran-order, and example file too. | 12:08 |
gsomix | sonne|work, aww :) ^ | 12:08 |
@wiking | .travis.yml file | 12:08 |
sonne|work | gsomix: then we should change that | 12:08 |
@wiking | since there has been some changes in the meanwhile | 12:08 |
@wiking | :P | 12:08 |
@wiking | sonne|work: ok so if there's ccache(-swig) then it's going to be enabled... | 12:09 |
sonne|work | cool | 12:09 |
@wiking | sonne|work: can i have access to a deb buildbot where there's enough ram where i could test python_modular interface? | 12:09 |
sonne|work | we don't have any :/ | 12:09 |
@wiking | :( | 12:10 |
@wiking | anybody can give me shell access to his machine which is a deb/ubuntu machine | 12:10 |
@wiking | plz | 12:10 |
gsomix | sonne|work, should we change data order in SGMatrix? O_o too many work, I think. | 12:10 |
van51 | wiking: sure | 12:11 |
@wiking | sonne|work: i'll add to cmake gmock and json in a way that if one wants it and the script cannot find a valid working one locally it'll try to download it itself | 12:11 |
@wiking | van51: u need a public key or something? | 12:11 |
sonne|work | gsomix: no not change it in SGMatrix | 12:11 |
sonne|work | wiking: woah maven! | 12:12 |
sonne|work | wiking: can you do that with eigen3 too? | 12:12 |
@wiking | sonne|work: yep | 12:12 |
@wiking | sonne|work: any package i can do | 12:12 |
sonne|work | but it should be possible to disable that right? | 12:12 |
@wiking | sonne|work: yeap i can do it in way | 12:13 |
@wiking | i mean in any way we wish | 12:13 |
sonne|work | wiking: sounds good | 12:13 |
van51 | wiking: I just set up an account for you | 12:13 |
@wiking | but i'll currently only set it up for json and gmock | 12:14 |
@wiking | sonne|work: since eigen has cmake as well... we can use that right away ;) | 12:14 |
@wiking | sonne|work: where does the buildbot master run? | 12:14 |
@wiking | sonne|work: and we should ask gunnar about that mac machine ;) | 12:16 |
@wiking | that he said it could be used for buildbot | 12:16 |
sonne|work | wiking: ohh good idea let me jsut do that | 12:16 |
@wiking | sonne|work: thnx | 12:17 |
sonne|work | errm food | 12:17 |
@wiking | van51: i'll try to do a build on your machine i hope you dont mind ;) | 12:19 |
gsomix | sonne|work, so, do you want just change example file? | 12:19 |
shogun-buildbot | build #1466 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1466 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 12:19 |
-!- travis-ci [~travis-ci@ec2-54-225-54-153.compute-1.amazonaws.com] has joined #shogun | 12:21 | |
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/9853318 | 12:21 |
-!- travis-ci [~travis-ci@ec2-54-225-54-153.compute-1.amazonaws.com] has left #shogun [] | 12:21 | |
@wiking | sonne|work: mmm we have a problem on travis... should we add some cmake variable so that a person can force to have python3 interface and not python2? | 12:24 |
* gsomix is gone for vodka. | 12:26 | |
@wiking | sonne|work: do u have access then to that cygwin machine? | 12:27 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has joined #shogun | 12:36 | |
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/9853659 | 12:36 |
-!- travis-ci [~travis-ci@ec2-184-73-123-75.compute-1.amazonaws.com] has left #shogun [] | 12:36 | |
@wiking | lol Shogun-2.2.0-Linux.deb | 12:40 |
@wiking | i have one like this ;) | 12:40 |
@wiking | very very cool | 12:42 |
@wiking | it works :) | 12:42 |
@wiking | although the package is not the way i would like to loook like | 12:42 |
shogun-buildbot | build #1161 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1161 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 12:43 |
-!- gsomix [~gsomix@178.45.37.229] has quit [Ping timeout: 240 seconds] | 12:45 | |
shogun-buildbot | build #843 of rpm1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/843 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 12:45 |
-!- gsomix [~gsomix@178.45.37.229] has joined #shogun | 12:46 | |
@wiking | van51: here? | 12:54 |
van51 | wiking: yes | 12:54 |
@wiking | van51: what's your distro on that machine? | 12:54 |
van51 | wiking: 13.04 | 12:54 |
van51 | wiking: ubuntu | 12:54 |
@wiking | van51: could u please install liboctave-dev for me ? | 12:55 |
van51 | wiking: sure | 12:55 |
van51 | wiking: done :) | 12:56 |
@wiking | ok great it works! | 12:56 |
van51 | wiking: I'm going for a swim, so I've given you sudo privileges in case you need anything else | 13:02 |
@wiking | van51: great thnx a lot! | 13:02 |
@wiking | but i hope now i'll have at least the octave done | 13:02 |
van51 | wiking: can you test it actually to make sure? | 13:02 |
@wiking | van51: well now i have a problem | 13:03 |
@wiking | so i'll debug it :) | 13:03 |
@wiking | /home/wiking/shogun/src/interfaces/octave_modular/sg_print_functions.cpp:1:27: fatal error: octave/config.h: No such file or directory | 13:03 |
@wiking | ;) | 13:03 |
van51 | wiking: ok work your magic :) | 13:03 |
van51 | wiking: but you do have sudo, right? | 13:04 |
@wiking | yep | 13:04 |
@wiking | works great | 13:04 |
@wiking | thnx | 13:04 |
van51 | wiking: np | 13:04 |
van51 | wiking: good luck :) | 13:04 |
van51 | and cu later | 13:04 |
@wiking | laterz | 13:04 |
van51 | sonne|work: btw I've sent a PR for the modular interfaces | 13:04 |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has quit [Quit: Leaving.] | 13:05 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has joined #shogun | 13:12 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 13:12 | |
shogun-notifier- | shogun: van51 :develop * 6d395aa / / (6 files): https://github.com/shogun-toolbox/shogun/commit/6d395aa1af961fb29af75fbf655885d971078bc4 | 13:13 |
shogun-notifier- | shogun: Added HashedDocDotFeatures in modular interfaces | 13:13 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 00bbcf7 / / (6 files): https://github.com/shogun-toolbox/shogun/commit/00bbcf7c63b4f6feda118771b3274cc82bdc0bea | 13:13 |
shogun-notifier- | shogun: Merge pull request #1363 from van51/feature/modular | 13:13 |
shogun-notifier- | shogun: | 13:13 |
shogun-notifier- | shogun: Added HashedDocDotFeatures in modular interfaces | 13:13 |
shogun-buildbot | build #844 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/844 | 13:26 |
sonne|work | wiking: herding your botnet ;) | 13:35 |
shogun-buildbot | build #1162 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1162 blamelist: van51 <vangelis_51@hotmail.com> | 13:39 |
shogun-buildbot | build #1163 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1163 blamelist: Soeren Sonnenburg <sonne@debian.org> | 13:40 |
shogun-buildbot | build #1468 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1468 blamelist: Soeren Sonnenburg <sonne@debian.org>, van51 <vangelis_51@hotmail.com> | 13:51 |
-!- travis-ci [~travis-ci@ec2-54-225-54-153.compute-1.amazonaws.com] has joined #shogun | 13:53 | |
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/9855223 | 13:53 |
-!- travis-ci [~travis-ci@ec2-54-225-54-153.compute-1.amazonaws.com] has left #shogun [] | 13:53 | |
shogun-buildbot | build #1467 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1467 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 14:38 |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has joined #shogun | 15:47 | |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has left #shogun [] | 15:49 | |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has joined #shogun | 15:51 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 16:13 | |
-!- iglesiasg [~Fernando@s83-179-44-135.cust.tele2.se] has quit [Quit: Leaving] | 16:23 | |
-!- thoralfk is now known as thoralf | 16:37 | |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:95d6:f5b6:4abf:728c] has joined #shogun | 16:57 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 16:57 | |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has joined #shogun | 17:13 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 17:14 | |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * d69a5fc / / (4 files): https://github.com/shogun-toolbox/shogun/commit/d69a5fc6ed335fd4948a57e3826862a1d7109ec4 | 17:14 |
shogun-notifier- | shogun: Automatically download Google testing framework if not available | 17:14 |
shogun-notifier- | shogun: Travis: attempt to fix python3 env | 17:14 |
@HeikoS | wiking: around? | 17:33 |
@HeikoS | or anyone: I need to run shogun on a computer where I dont have make install rights | 17:34 |
@HeikoS | how? | 17:34 |
@HeikoS | I compiled with prefix=bla but I cannot do make install anyway | 17:34 |
@HeikoS | I want to use the python modular | 17:34 |
pickle27 | hmmmm | 17:34 |
pickle27 | can you copy the lib file to your working python dir? I don't even know if that would work | 17:35 |
@HeikoS | pickle27: yeah I need something like that :) | 17:36 |
thoralf | HeikoS: You have to provide the install paths with configure. | 17:36 |
@HeikoS | but I want to avoid doing hacks ;) | 17:36 |
@HeikoS | thoralf: you mean prefix? | 17:37 |
thoralf | HeikoS: No. | 17:37 |
@HeikoS | thoralf: problem is that I cannot do make install | 17:37 |
@HeikoS | the make installation doesnt allow it | 17:37 |
thoralf | --pydir=DIR | 17:37 |
thoralf | HeikoS: It's not automatically set correctly whet giving --prefix. | 17:37 |
thoralf | HeikoS: Clearly a but. | 17:38 |
-!- travis-ci [~travis-ci@ec2-23-22-133-41.compute-1.amazonaws.com] has joined #shogun | 17:38 | |
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/9863553 | 17:38 |
-!- travis-ci [~travis-ci@ec2-23-22-133-41.compute-1.amazonaws.com] has left #shogun [] | 17:38 | |
thoralf | bug | 17:38 |
@HeikoS | thoralf: Sorry user bla is not allowed to execute /usr/bin/make install as root | 17:38 |
thoralf | HeikoS: Have a look into .config, parameter PYDIR | 17:38 |
@HeikoS | thoralf: ah, I just did it wothout root ;) | 17:38 |
@HeikoS | pydir is the target of the python installation? | 17:38 |
thoralf | HeikoS: Yes. | 17:39 |
@HeikoS | thoralf: cool, where do I have to let that point? | 17:39 |
thoralf | HeikoS: I don't understand. What do you mean? | 17:39 |
@sonney2k | HeikoS, errm it works | 17:40 |
@HeikoS | thoralf: so I now installed it locally to somedir/lib/python2.7 | 17:40 |
@HeikoS | sonney2k: yeah I just never did this and asking how :) | 17:40 |
@sonney2k | HeikoS, just do ./configure --prefix=~/my/local/path | 17:40 |
@HeikoS | how to call python so that it recognises the installation | 17:40 |
@HeikoS | sonney2k: did that | 17:40 |
@HeikoS | and the? | 17:40 |
@sonney2k | then make install | 17:40 |
@HeikoS | did | 17:40 |
@HeikoS | and then? | 17:40 |
@HeikoS | add pythonpath? | 17:41 |
@sonney2k | PYTHONPATH=$HOME/my/local/path/deep/down/there LD_LIBRARY_PATH=$HOME/my/local/path/lib python | 17:41 |
@sonney2k | import modshogun | 17:41 |
@sonney2k | from python then | 17:41 |
@HeikoS | sonney2k: ah that easy, cool thanks! | 17:41 |
@sonney2k | HeikoS, the PYTHONPATH has to point to the dir where you find modshogun | 17:41 |
@sonney2k | HeikoS, lisitsyn: btw do you have a minute? I want to change behavior of ascii readers quite a bit. | 17:42 |
@sonney2k | to make each line an example | 17:42 |
@sonney2k | e.g. | 17:42 |
@HeikoS | sonney2k: yeah on one ear | 17:42 |
@sonney2k | for strings | 17:42 |
@sonney2k | 'ablfjdslkfjaslkfasdjf\n' | 17:43 |
@HeikoS | sonney2k: I never used any of this stuff so far | 17:43 |
@HeikoS | always using python for that | 17:43 |
@sonney2k | would be in one line -> one string later in shogun | 17:43 |
@HeikoS | sonney2k: sounds good | 17:43 |
thoralf | sonney2k: --prefix does not work for octave, btw. | 17:43 |
@sonney2k | for sparse features same thing: one line -> one sparse feature vector | 17:43 |
thoralf | sonney2k: The path is still global. | 17:43 |
@sonney2k | thoralf, I know - octave doesn't support this | 17:43 |
@sonney2k | thoralf, at least 2.x / 3.0 when I tried to add support for that | 17:44 |
@sonney2k | thoralf, if it works now tell my how! | 17:44 |
@sonney2k | HeikoS, well that already works | 17:44 |
@sonney2k | HeikoS, problem is now for dense matrices | 17:44 |
@sonney2k | this will lead to a transposed view in SGMatrix | 17:44 |
@sonney2k | because one line -> one column vector! | 17:44 |
@sonney2k | becauase one column vector == one example in shogun | 17:45 |
@sonney2k | and this might confuse people | 17:45 |
@HeikoS | sonney2k: ah I see | 17:45 |
@HeikoS | sonney2k: btw any reason for the col format matrices? | 17:45 |
@sonney2k | and FWIW octave and matlab write matrices not as I would do it | 17:46 |
@sonney2k | HeikoS, yes sure | 17:46 |
@HeikoS | sonney2k: it would be so cool if we had numpy style views on matrices | 17:46 |
@sonney2k | HeikoS, simply if you want just 100 examples | 17:46 |
@HeikoS | with slice operations etc | 17:46 |
@sonney2k | you can stop reading after 100 lines | 17:46 |
@sonney2k | and you can do streaming from the same code | 17:46 |
@HeikoS | sonney2k: ok makes sense | 17:46 |
* thoralf is already using steramingasciifile for this and creates sparse matrices ;) | 17:46 | |
@sonney2k | and it is more consistent considering sparse/string stuff where one line == one example | 17:46 |
@sonney2k | soon ascii file will die :D | 17:47 |
thoralf | sonney2k: Oh. | 17:47 |
@sonney2k | gsomix' CSVFile is replacing it | 17:47 |
@sonney2k | it already does everything asciifile does but faster and with configurable seperators | 17:47 |
@sonney2k | HeikoS, to do this I will need to transpose our dense real toy data with transposed versions | 17:48 |
thoralf | sonney2k: As long as it's supported by streaming sparse features, it's not a big deal. | 17:48 |
thoralf | sonney2k: Or something similar, if this dies, too. ;) | 17:49 |
@sonney2k | thoralf, yes it will :D | 17:49 |
@HeikoS | sonney2k: ah I see | 17:49 |
@HeikoS | sonney2k: bit I agree on it | 17:49 |
thoralf | sonney2k: It will die or it will support ssf? ;) | 17:50 |
@sonney2k | thoralf, CSV will do it | 17:50 |
@sonney2k | HeikoS, anyway this is only relevant for csv data - in fact with csv users usually have 1 example == 1 line | 17:51 |
@sonney2k | so IMHO it would be confusing not to do it | 17:51 |
lisitsyn | sonney2k: here | 17:51 |
@sonney2k | we could have an option to read the transpose (.tsv) | 17:51 |
@sonney2k | lisitsyn, would you mind having one line == one example notion also for ascii dense matrices | 17:52 |
lisitsyn | sonney2k: no don't mind at all | 17:52 |
lisitsyn | transpose option is the best though I guess? | 17:52 |
@sonney2k | lisitsyn, gsomix' reader already supports fortran and c order | 17:52 |
gsomix | hey guys | 17:52 |
@sonney2k | so just specifying fortran order would make the transpose loaded | 17:53 |
gsomix | I have some vodka - time to work | 17:53 |
@sonney2k | gsomix, what are you working on now btw? | 17:53 |
@sonney2k | I mean I guess I will know soon from your email :D | 17:53 |
gsomix | sonney2k, still libsvm file/quoting/refactoring for CircularBuffer. I'll send PR today. | 17:55 |
gsomix | I hope vodka helps me | 17:55 |
lisitsyn | sonney2k: carol said the waiting list opens september 6 | 17:55 |
lisitsyn | I am out :) | 17:55 |
lisitsyn | time to go home | 17:56 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Quit: Leaving.] | 17:56 | |
thoralf | Aww. This swig thing only uses one core to compile... taking ages. :( | 17:56 |
thoralf | sonney2k: I guess octave should work from local directory, I'm about to check it, but need to wait for swig. | 17:57 |
@sonney2k | thoralf, yeah if you have an idea how to split a .cpp file into 10 or so it would be nice to compile that all in parallel | 17:57 |
@sonney2k | thoralf, yes but where do I copy the sg.so / modshogun.so then? | 17:58 |
thoralf | sonney2k: It's one big cpp file? Jesus. ;) | 17:58 |
@sonney2k | thoralf, yes | 17:59 |
@sonney2k | gtg | 17:59 |
@sonney2k | cu | 17:59 |
thoralf | sonney2k: octave_modular/_static already work when calling "make tests-local". So it's already set up properly. | 18:00 |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * 652e041 / .travis.yml,CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/652e0416f65a3dca253a629d707ca751fb8224d2 | 18:11 |
shogun-notifier- | shogun: Travis: attempt to fix ruby_modular interface | 18:11 |
shogun-notifier- | shogun: remove on failure ctest setting | 18:11 |
-!- hoijui [~hoijui@adsl-84-227-37-53.adslplus.ch] has joined #shogun | 18:49 | |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has joined #shogun | 19:46 | |
shogun-notifier- | shogun: van51 :develop * d21ef9a / / (3 files): https://github.com/shogun-toolbox/shogun/commit/d21ef9a8945c58e71837eb58c7f75a31a82c8985 | 19:50 |
shogun-notifier- | shogun: Added HashedDocConverter to modular interfaces | 19:50 |
shogun-notifier- | shogun: Soeren Sonnenburg :develop * 74d2c7d / / (3 files): https://github.com/shogun-toolbox/shogun/commit/74d2c7d468aaf218a35ff7b86f10774a53ca6688 | 19:50 |
shogun-notifier- | shogun: Merge pull request #1365 from van51/feature/modular | 19:50 |
shogun-notifier- | shogun: | 19:50 |
shogun-notifier- | shogun: Added HashedDocConverter to modular interfaces | 19:50 |
-!- lambday [67157c37@gateway/web/freenode/ip.103.21.124.55] has quit [Ping timeout: 250 seconds] | 19:54 | |
-!- lisitsyn [~lisitsyn@213.87.136.83] has joined #shogun | 19:59 | |
shogun-buildbot | build #1164 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1164 blamelist: van51 <vangelis_51@hotmail.com> | 20:09 |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has quit [Quit: Leaving.] | 20:15 | |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has joined #shogun | 20:21 | |
@iglesiasg | hushell, I am heading home now. I will have dinner and then read your PR! | 20:27 |
-!- iglesiasg [~iglesias@2001:6b0:1:1da0:95d6:f5b6:4abf:728c] has quit [Quit: Ex-Chat] | 20:28 | |
shogun-buildbot | build #1469 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1469 blamelist: van51 <vangelis_51@hotmail.com> | 20:28 |
-!- travis-ci [~travis-ci@ec2-23-22-133-41.compute-1.amazonaws.com] has joined #shogun | 20:34 | |
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/9869262 | 20:34 |
-!- travis-ci [~travis-ci@ec2-23-22-133-41.compute-1.amazonaws.com] has left #shogun [] | 20:34 | |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * 610c9cf / CMakeLists.txt,src/shogun/lib/config.h.in: https://github.com/shogun-toolbox/shogun/commit/610c9cf53effe388852070b13251624dbdc90387 | 20:50 |
shogun-notifier- | shogun: Add ccache support (ccache-swig is still missing) | 20:50 |
shogun-notifier- | shogun: Set COMPFLAGS_CPP in config.h and add HAVE_CXX macro to config.h | 20:50 |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * 4c6865b / cmake/FindCCache.cmake: https://github.com/shogun-toolbox/shogun/commit/4c6865bb2ccab1d1146b5554908d9098f4d092c8 | 20:51 |
shogun-notifier- | shogun: Add FindCCache.cmake script for detecting ccache | 20:51 |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 256 seconds] | 20:54 | |
foulwall | ping lisitsyn sonne|work | 20:58 |
lisitsyn | foulwall: hey | 20:58 |
foulwall | lisitsyn: I've make k alterable on word embedding | 20:59 |
foulwall | lisitsyn: have a look at http://nn.7nn.de:8000/tapkee/entrance | 20:59 |
foulwall | and select feature type 'words_embedding' | 21:00 |
shogun-buildbot | build #1165 of cyg1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1165 | 21:00 |
foulwall | kinda slow | 21:00 |
lisitsyn | should be pretty fast hmm | 21:00 |
shogun-buildbot | build #1470 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1470 blamelist: Soeren Sonnenburg <sonne@debian.org> | 21:00 |
foulwall | lisitsyn: I did dimred in this way, https://github.com/ZhengyangL/shogun-demo/blob/master/demo/tapkee.py | 21:02 |
-!- hoijui [~hoijui@adsl-84-227-37-53.adslplus.ch] has quit [Ping timeout: 264 seconds] | 21:07 | |
-!- gsomix_ [~gsomix@178.45.86.56] has joined #shogun | 21:21 | |
-!- gsomix [~gsomix@178.45.37.229] has quit [Read error: Connection reset by peer] | 21:21 | |
@wiking | sonney2k: ping | 21:30 |
@wiking | lisitsyn: Shogun-2.2.0-Linux-libraries.deb | 21:30 |
@wiking | heheh | 21:30 |
lisitsyn | wiking: haha nice | 21:31 |
@wiking | although the naming is still wrong :S | 21:32 |
@wiking | and need to set the dependencies | 21:32 |
@wiking | but in the meanwhile ccache is already automaticaly set + gmock/gtest is autofetched from the website if not present :P | 21:33 |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has quit [Read error: Connection reset by peer] | 21:43 | |
-!- travis-ci [~travis-ci@ec2-50-16-141-3.compute-1.amazonaws.com] has joined #shogun | 21:44 | |
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/9871676 | 21:44 |
-!- travis-ci [~travis-ci@ec2-50-16-141-3.compute-1.amazonaws.com] has left #shogun [] | 21:44 | |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has joined #shogun | 21:45 | |
@HeikoS | lisitsyn: around? | 21:59 |
@HeikoS | [ RUN ] IsomapTest.neighbors_preserving | 22:00 |
@HeikoS | Segmentation fault (core dumped) | 22:00 |
lisitsyn | HeikoS: yes | 22:00 |
@HeikoS | lisitsyn: do you have that too? | 22:00 |
lisitsyn | cool | 22:01 |
lisitsyn | lets see if I reproduce it | 22:01 |
van51 | sonney2k: I've updated the PR for the hashed doc features. I think it's OK now | 22:02 |
lisitsyn | HeikoS: idk checking | 22:02 |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * 7409cc0 / CMakeLists.txt/ (3 files): https://github.com/shogun-toolbox/shogun/commit/7409cc013e2d73868103141ba03ade14d1dd1e3b | 22:03 |
shogun-notifier- | shogun: Install libshogun examples under share/shogun/examples/libshogun | 22:03 |
shogun-notifier- | shogun: add initial component based packaging for DEB and RPM | 22:03 |
@wiking | yey motherfuckers | 22:03 |
@HeikoS | wiking: whoooo | 22:03 |
@wiking | ok so now i'm still missing ccache-swig support | 22:04 |
@wiking | and the fucking travis builds for octave, ruby and python3 | 22:04 |
@wiking | i cannot find the fucking travis vm image anywhere for download | 22:04 |
@wiking | ah and btw: no _static interface is supported at all :D | 22:05 |
@HeikoS | wiking: ha! | 22:06 |
@HeikoS | that might be good :) | 22:06 |
-!- van51 [~van51@ppp-94-66-42-68.home.otenet.gr] has quit [Quit: Leaving.] | 22:07 | |
@HeikoS | wiking: I have made some progress with the notebooks btw | 22:07 |
@HeikoS | think I will get it working soon | 22:07 |
@wiking | that's great | 22:07 |
@wiking | i'll have to add it to the cmake :P | 22:07 |
@HeikoS | wiking: nice :) | 22:07 |
@HeikoS | wiking: btw do you have any ideas for the docsprint? should we do that? | 22:07 |
@HeikoS | wiking: still one day left | 22:07 |
@HeikoS | until deadline | 22:08 |
@wiking | HeikoS: heheheh :D | 22:08 |
@wiking | is it already? :) | 22:08 |
@HeikoS | wiking: yeah | 22:08 |
@wiking | fuck | 22:08 |
@HeikoS | I am terribly busy unitl tomorrow unfortunately but maybe we should submit something | 22:08 |
@HeikoS | but just copying last year's application wont do it since we were rejected | 22:08 |
@HeikoS | we need a "vision" for a book | 22:08 |
@HeikoS | and maybe some power users, like thoralf | 22:09 |
@HeikoS | and I think just a list of ML algorithms also wont do it | 22:09 |
@HeikoS | should be less scientific, more cummunity based | 22:09 |
@HeikoS | but dont know | 22:09 |
@HeikoS | whether its worth the hassle, last year, kind of big organisations were picked, like KDE | 22:09 |
@wiking | ehehe | 22:10 |
@HeikoS | wiking: so what do you think? | 22:10 |
@wiking | mmm worth a try imho | 22:11 |
@wiking | if nothing else the copy paste | 22:11 |
@wiking | i mean why not :) | 22:11 |
@wiking | worst case we get another reject | 22:11 |
@wiking | nothing bad can come out of that | 22:11 |
@HeikoS | wiking: ah | 22:11 |
@HeikoS | wow | 22:11 |
@HeikoS | deadline just got extended :) | 22:11 |
@HeikoS | 2 mins ago | 22:11 |
@wiking | heheh | 22:11 |
@HeikoS | yeah! :) | 22:11 |
@HeikoS | then I can do it end of the week | 22:11 |
@HeikoS | phew | 22:11 |
@wiking | which means there's not enough | 22:11 |
@wiking | applications | 22:12 |
@HeikoS | The response from individuals has been tremendous, but we'd like to encourage more projects to apply as well. | 22:12 |
@wiking | which is good for us | 22:12 |
@HeikoS | lets doooo it then | 22:12 |
@wiking | yeah lets do it | 22:12 |
@HeikoS | wiking: cool ill take care after tues | 22:12 |
shogun-notifier- | shogun: Viktor Gal :feature/CMake * 1c40b9e / .travis.yml: https://github.com/shogun-toolbox/shogun/commit/1c40b9ed1cb2e53658c629ae78cb8f4d84dab973 | 22:17 |
shogun-notifier- | shogun: Travis: fix python3 and ruby modular (attempt 3) | 22:17 |
@wiking | attempt number | 22:18 |
@wiking | ... | 22:18 |
@wiking | i hope travis builds starts to work finalluy | 22:18 |
-!- travis-ci [~travis-ci@ec2-50-16-141-3.compute-1.amazonaws.com] has joined #shogun | 22:19 | |
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/9871731 | 22:19 |
-!- travis-ci [~travis-ci@ec2-50-16-141-3.compute-1.amazonaws.com] has left #shogun [] | 22:19 | |
lisitsyn | HeikoS: it is ok here | 22:25 |
lisitsyn | could you provide something more? | 22:25 |
-!- iglesiasg [~iglesias@s83-179-44-135.cust.tele2.se] has joined #shogun | 22:27 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 22:27 | |
@HeikoS | lisitsyn: make clean helps | 22:35 |
lisitsyn | HeikoS: no idea what it was | 22:36 |
@HeikoS | lisitsyn, wiking is there any way to detect this make clean weirdness | 22:36 |
@HeikoS | recompiling libshogun often gives crashes in unit tests | 22:36 |
@HeikoS | make clean;make solves it | 22:36 |
@HeikoS | would be nice tobe detected automatically | 22:36 |
lisitsyn | HeikoS: I think cmake is better in this | 22:36 |
@HeikoS | ok | 22:36 |
lisitsyn | but not sure | 22:36 |
lisitsyn | well with cmake we rely on cmake | 22:37 |
@HeikoS | lisitsyn, wiking another question: | 22:38 |
@HeikoS | here at uni I have the problem that there are many computers with different architectures | 22:38 |
@HeikoS | so when I run stuff on the cluster, I cannot use my locally compiled shogun installation | 22:38 |
@HeikoS | but would have to create one for every system and match them | 22:39 |
@HeikoS | any ideas how to solve this best? | 22:39 |
@HeikoS | also, I cannot guarantee the installed packages on the computers | 22:40 |
@HeikoS | will cmake help me with this? :) | 22:40 |
lisitsyn | HeikoS: haha | 22:40 |
lisitsyn | will cmake cure cancer | 22:40 |
@wiking | some modtherfucker did this | 22:40 |
@HeikoS | I would like to just add a one-liner to my scripts that installs shogun locally for the given architecture | 22:41 |
@wiking | javac] /home/travis/build/shogun-toolbox/shogun/src/interfaces/java_modular/org/shogun/CProbabilityDistribution.java:111: error: method log_pdf(Doublatrix) is already defined in class CProbabilityDistribution | 22:41 |
@wiking | [javac] public double log_pdf(Doublatrix single_sample) { | 22:41 |
@wiking | [javac] ^ | 22:41 |
@wiking | [javac] /home/travis/build/shogun-toolbox/shogun/src/interfaces/java_modular/org/shogun/LikelihoodModel.java:127: error: method get_log_probability_f(Labels,Doublatrix) is already defined in class LikelihoodModel | 22:41 |
@wiking | [javac] public Doublatrix get_log_probability_f(Labels lab, Doublatrix F) { | 22:41 |
@wiking | [javac] ^ | 22:41 |
lisitsyn | wiking: I was looking at it and saw no error here | 22:41 |
@wiking | [javac] 2 errors | 22:41 |
lisitsyn | Doublatrix? | 22:41 |
@wiking | lisitsyn: that's cmake error | 22:41 |
lisitsyn | what? | 22:41 |
@HeikoS | wiking: that probably was me, what is that? | 22:41 |
lisitsyn | Doublatrix? | 22:41 |
@wiking | sometimes it eats some chars | 22:41 |
@wiking | the console jquery is sometimes ... | 22:41 |
@wiking | so it's shit | 22:42 |
lisitsyn | ahh | 22:42 |
@wiking | but dont worry it's not that | 22:42 |
lisitsyn | so that's not | 22:42 |
lisitsyn | doubltrix | 22:42 |
lisitsyn | oh | 22:42 |
@wiking | nop | 22:42 |
lisitsyn | phew | 22:42 |
lisitsyn | :D | 22:42 |
@HeikoS | virtual SGVector<float64_t> log_pdf(SGMatrix<float64_t> samples) const; | 22:42 |
@HeikoS | virtual float64_t log_pdf(SGVector<float64_t> single_sample) const; | 22:42 |
@HeikoS | is this the problem? | 22:42 |
lisitsyn | ahhhhh | 22:42 |
@wiking | yes | 22:42 |
lisitsyn | yes of course | 22:42 |
@HeikoS | both of those methods are defined | 22:42 |
@wiking | it's a major problems | 22:42 |
lisitsyn | but return value | 22:42 |
@HeikoS | what? | 22:42 |
@wiking | u cannot have that | 22:42 |
@HeikoS | why? | 22:42 |
@wiking | because c++ | 22:42 |
lisitsyn | HeikoS: SGVector = SGMatrix | 22:42 |
lisitsyn | in java | 22:42 |
lisitsyn | no that's not because C++ | 22:42 |
@wiking | lisitsyn: noup | 22:42 |
lisitsyn | C++ is ok here | 22:43 |
@HeikoS | oh thats stupid | 22:43 |
@wiking | why? | 22:43 |
@wiking | ah yeah | 22:43 |
@wiking | ok | 22:43 |
@wiking | i see | 22:43 |
@HeikoS | I would love them to have a common base | 22:43 |
lisitsyn | wiking: everything is ok C++ wise | 22:43 |
@HeikoS | I think its a restriction | 22:43 |
@wiking | lisitsyn: yeah misinterpreted the func arg in my had | 22:43 |
lisitsyn | HeikoS: jblas doesn't have such stuff | 22:43 |
@wiking | *head | 22:43 |
@HeikoS | but now I cannot even rely on this | 22:43 |
@HeikoS | mmh | 22:43 |
@HeikoS | so rename? | 22:43 |
lisitsyn | jblas has just matrix | 22:43 |
lisitsyn | yes rename will help here | 22:43 |
@wiking | HeikoS: i wonder if u can rename it only at modular level :P | 22:44 |
@wiking | but that would be misleading | 22:44 |
@wiking | soooo | 22:44 |
@wiking | go for a full rename | 22:44 |
@HeikoS | ok then | 22:44 |
@wiking | life's a bitch my friend | 22:44 |
@wiking | ;P | 22:44 |
@wiking | ain't no likey the stuff :D | 22:44 |
@HeikoS | wiking: did you see my question above? any ideas? | 22:44 |
@wiking | HeikoS: mmmm welll | 22:44 |
@wiking | HeikoS: u'll have to do it one by one the build | 22:45 |
@wiking | HeikoS: or be a hero and create the ExternalProject_Add | 22:45 |
@wiking | for all the dependencies in shogun | 22:45 |
@wiking | cmake... | 22:45 |
@wiking | that'll fetch all the fucking dependency and build it with shogun | 22:45 |
@wiking | all your problems will be gone | 22:45 |
@wiking | ;) | 22:45 |
@wiking | but maybe your compile time will be a bit big | 22:46 |
@HeikoS | wiking: this would compile all dependencies right? | 22:46 |
@wiking | HeikoS: y | 22:46 |
@HeikoS | wiking: mmmh, so on some machines I cannot even install eigen | 22:46 |
@HeikoS | what about half half | 22:46 |
@wiking | HeikoS: u dont care about that part | 22:46 |
@HeikoS | python and swig is almost everywhere | 22:47 |
@wiking | HeikoS: you'll build all the deps internally that are not found on the system | 22:47 |
@HeikoS | wiking: yeah | 22:47 |
@HeikoS | that sounds good | 22:47 |
@wiking | HeikoS: this can be done with cmake | 22:47 |
@HeikoS | and then install locally and add to path bla | 22:47 |
@HeikoS | wiking: nice | 22:47 |
@wiking | HeikoS: but one needs to create this | 22:47 |
@HeikoS | wonderful, will make my life so much easier | 22:47 |
@wiking | HeikoS: no you dont need to add to path anything ... | 22:47 |
@wiking | cmake will take care of it | 22:47 |
@wiking | but | 22:47 |
@wiking | one has to write these | 22:47 |
@wiking | for all the fucking libraries we are depending on | 22:48 |
@wiking | which is quite seom | 22:48 |
@wiking | *some | 22:48 |
@HeikoS | wiking: how and where? I can do that | 22:48 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/feature/CMake/tests/unit/CMakeLists.txt | 22:49 |
@wiking | HeikoS: there's an example for ya | 22:49 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/feature/CMake/tests/unit/CMakeLists.txt#L10 | 22:49 |
@wiking | to be precise | 22:49 |
@wiking | it downloads and compiles gtest/gmock | 22:49 |
@HeikoS | should I modify this file? | 22:50 |
@wiking | no | 22:50 |
lisitsyn | no that's tests cmake lists | 22:50 |
@wiking | that's only an exmaple | 22:50 |
@wiking | what u should do is | 22:50 |
@wiking | first think about a general way how to do this | 22:50 |
lisitsyn | HeikoS: have you worked with maven? | 22:50 |
@wiking | but i would say | 22:51 |
@wiking | in the cmake dir | 22:51 |
@wiking | we should add a new dir called | 22:51 |
@wiking | cmake/external | 22:51 |
@wiking | and start adding there the cmake scripts for the various libraries we depend on | 22:51 |
lisitsyn | we can distribite eigen3 along with our sources actually | 22:51 |
@wiking | and then we can include those wherever we need them | 22:51 |
lisitsyn | just headers | 22:51 |
lisitsyn | :D | 22:51 |
@wiking | lisitsyn: yes | 22:52 |
lisitsyn | but IIRC sonney2k is not a big fan of it | 22:52 |
@wiking | but you can just compile it | 22:52 |
@wiking | and not distribute | 22:52 |
@wiking | and that's good enough imho | 22:52 |
lisitsyn | compile what? | 22:52 |
@sonney2k | wiking, pong | 22:52 |
@wiking | lisitsyn: just simply have eigen3 as a third_party and then compile the libshogun with that eigen3 | 22:52 |
lisitsyn | oh I summoned sonney2k | 22:52 |
@wiking | sonney2k: mmm i have no idea what i wanted :) | 22:53 |
@wiking | sonney2k: oh yeah i have deb packages for libshogun | 22:53 |
@wiking | ;) | 22:53 |
@wiking | sonney2k: you should start thinking about where to upload them | 22:53 |
shogun-notifier- | shogun: Heiko Strathmann :develop * dcec885 / src/shogun/base/SGObject.cpp: https://github.com/shogun-toolbox/shogun/commit/dcec8853d9682bac126d04cb7c1babe59d3d0a49 | 22:53 |
shogun-notifier- | shogun: added a debug message on current parameter processed | 22:53 |
shogun-notifier- | shogun: Heiko Strathmann :develop * f525112 / src/shogun/labels/DenseLabels.cpp,src/shogun/labels/DenseLabels.h: https://github.com/shogun-toolbox/shogun/commit/f5251120f09337b9b54c8b3d4161a98136ab6f04 | 22:53 |
shogun-notifier- | shogun: make get_labels work with a subset (create copy) | 22:53 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 6991f40 / src/shogun/machine/GaussianProcessMachine.h: https://github.com/shogun-toolbox/shogun/commit/6991f40149942499af44ab32a83e93da6e83a965 | 22:53 |
shogun-notifier- | shogun: add empty experimental method to try cross-validation with GPs | 22:53 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 38b3615 / src/NEWS: https://github.com/shogun-toolbox/shogun/commit/38b3615c19835cae0b729e3ff97f9d67ca8095f0 | 22:53 |
shogun-notifier- | shogun: add note about (soon to be working) cross-validation with GPs | 22:53 |
shogun-notifier- | shogun: Heiko Strathmann :develop * d968f7d / / (6 files): https://github.com/shogun-toolbox/shogun/commit/d968f7d0b4acdc2ff5076bc17c88455aa0608740 | 22:53 |
shogun-notifier- | shogun: rename method to avoid clash in java interface where SGMatrix==SGVector | 22:53 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 8935269 / / (11 files): https://github.com/shogun-toolbox/shogun/commit/89352693b26084613ce993705bb7b07249e0f6d5 | 22:53 |
shogun-notifier- | shogun: Merge pull request #1366 from karlnapf/develop | 22:53 |
shogun-notifier- | shogun: | 22:53 |
shogun-notifier- | shogun: Some minor changes towards x-validation for GPC | 22:53 |
@wiking | sonney2k: and as well what the fuck is the naming convention for nightly built deb packages? | 22:53 |
@wiking | i mean for version | 22:54 |
@wiking | the git string? | 22:54 |
@sonney2k | wiking, we could create an apt archive thing on shogun-toolbox.org | 22:55 |
@wiking | yes | 22:55 |
@sonney2k | wiking, there is no clear naming convention | 22:55 |
@sonney2k | wiking, some people use 1.2.0+git<hash> | 22:56 |
@wiking | ok | 22:56 |
@wiking | then it'll be something like that | 22:56 |
@HeikoS | lisitsyn: i used maven like 4 years ago for a bit at uni, but forgot everything | 22:56 |
@sonney2k | some use 1.2.0+20130805git<hash>-1 | 22:57 |
@sonney2k | wiking, btw I am currently setting up our team address - should it be shogun-team@ ... ? | 22:59 |
@sonney2k | some mailman mailinglist | 22:59 |
@wiking | mmm yeah something like that | 23:00 |
@wiking | sonney2k: i mean even an alias is good | 23:01 |
lisitsyn | HeikoS: that's sth like maven then ;) | 23:01 |
@sonney2k | wiking, list is easiest we have the infrastructure already and so should be using it | 23:03 |
@sonney2k | lisitsyn, why is the words embedding so slow? | 23:04 |
lisitsyn | sonney2k: no idea | 23:05 |
@wiking | fucking us is awake... travis is much slower at these times | 23:07 |
shogun-buildbot | build #1166 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1166 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 23:10 |
shogun-buildbot | build #1167 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1167 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 23:11 |
@wiking | sonney2k: so what's with this cygwin machine? | 23:13 |
@wiking | sonney2k: can u install a VS on it? | 23:13 |
@iglesiasg | hushell, hey!, you there? | 23:17 |
-!- hushell [~hushell@c-24-21-169-136.hsd1.or.comcast.net] has quit [Quit: WeeChat 0.3.7] | 23:22 | |
@sonney2k | wiking, didn't get a reply | 23:27 |
@iglesiasg | shogun-team rocks! | 23:34 |
@sonney2k | iglesiasg, HeikoS, wiking , lisitsyn - it is basically us 5 on this list | 23:34 |
@sonney2k | lets create a youtube account | 23:35 |
lisitsyn | what is it for? | 23:35 |
lisitsyn | ahh | 23:35 |
shogun-buildbot | build #1471 of deb3 - modular_interfaces is complete: Failure [failed compile java_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/1471 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 23:36 |
@HeikoS | we should have dev meetings twice a year :) | 23:42 |
@sonney2k | wiking, feel free to start uploading the videos! | 23:42 |
@wiking | sonney2k: thnx | 23:42 |
lisitsyn | HeikoS: I am waiting you here in january | 23:42 |
@wiking | sonney2k: i'll need uplink for that | 23:42 |
lisitsyn | :D | 23:42 |
@sonney2k | HeikoS, we should maybe have some monthly(?) irc / google hangout based discussions | 23:43 |
@HeikoS | no but seriously, would be very cool I think | 23:43 |
@HeikoS | sonney2k: yep I also agree | 23:43 |
@sonney2k | I mean we have lots to discuss but we never take the time | 23:43 |
@HeikoS | yep I agree | 23:43 |
@HeikoS | sonney2k: one solution would be to discuss in the issues | 23:43 |
@sonney2k | so better have some evening every no and then and do some decision in a *focussed* way | 23:43 |
@HeikoS | That was my idea with this general ideas to discuss | 23:44 |
@iglesiasg | I think it is a good idea to do these meetings | 23:44 |
@sonney2k | wiking, uplink? not on your server then I guess? | 23:44 |
@HeikoS | sonney2k: and then paste the logs somewhere | 23:44 |
@HeikoS | or even somebody has to summarise | 23:44 |
@HeikoS | so that we and others know whats going on | 23:44 |
@HeikoS | thats actually a nice thing | 23:44 |
@sonney2k | HeikoS, or IRC... etherpad etc | 23:44 |
@HeikoS | or even google hangouts :) | 23:45 |
@HeikoS | if we are <5 | 23:45 |
lisitsyn | they should called it google hangover | 23:45 |
-!- hushell [~hushell@c-24-21-169-136.hsd1.or.comcast.net] has joined #shogun | 23:46 | |
@wiking | lol | 23:48 |
@wiking | /root/shogun/src/shogun/preprocessor/DensePreprocessor.cpp:1:0: error: CPU you selected does not support x86-64 instruction set | 23:48 |
-!- pickle27 [~Kevin@d67-193-243-174.home3.cgocable.net] has quit [Quit: Leaving] | 23:48 | |
@wiking | how great to have shogun in qemu image :) | 23:48 |
@sonney2k | wiking, haha | 23:49 |
@wiking | i wonder what could be the fix for this | 23:49 |
@sonney2k | wiking, I usually compile just with -O3 but no -march etc flags | 23:49 |
@sonney2k | then it will work | 23:49 |
@sonney2k | a lot slower for sure but it will work | 23:50 |
@wiking | yeye but still... | 23:50 |
@wiking | i want to be able to build this shit in qemu | 23:50 |
@wiking | with cmake | 23:50 |
@wiking | so fuck it | 23:50 |
@iglesiasg | sonney2k, about java_modular | 23:51 |
@iglesiasg | virtual SGVector<float64_t> log_pdf(SGMatrix<float64_t> samples) const; | 23:51 |
@iglesiasg | this one together with this other one | 23:51 |
@iglesiasg | virtual float64_t log_pdf(SGVector<float64_t> single_sample) const; | 23:52 |
@iglesiasg | give a compile error | 23:52 |
@iglesiasg | I checked the .java, and it is because both SGVector and SGMatrix are traduced to DoubleMatrix in java | 23:52 |
@iglesiasg | but I have seen some RealVector in other .java created by SWIG | 23:53 |
@iglesiasg | one of them was a SGVector<float64_t>& | 23:53 |
hushell | iglesiasg: Hey! Thanks the review! | 23:55 |
@iglesiasg | hushell, you are welcome :) | 23:55 |
hushell | iglesiasg: any idea why MAPInference::clone() failed? | 23:55 |
@iglesiasg | hushell, let me check the class better | 23:56 |
-!- lisitsyn [~lisitsyn@213.87.136.83] has quit [Quit: Leaving.] | 23:56 | |
hushell | iglesiasg: http://pastebin.com/8kKKjJJw | 23:57 |
shogun-buildbot | build #1168 of cyg1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/1168 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 23:57 |
hushell | iglesiasg: this is my test code | 23:57 |
@iglesiasg | hushell, segmentation fault? | 23:58 |
@HeikoS | sonney2k: btw I thought about this wiki vs readme issue | 23:58 |
@HeikoS | sonney2k: someone told me its possible to generate website documentation from a readme with github .... | 23:59 |
@iglesiasg | HeikoS, sure, it is super easy. gh-pages branch | 23:59 |
@iglesiasg | HeikoS, http://iglesias.github.io/linal/ | 23:59 |
--- Log closed Tue Aug 06 00:00:00 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!