--- Log opened Sun Jul 01 00:00:17 2012 | ||
-!- blackburn [~blackburn@81.28.187.0] has quit [Ping timeout: 248 seconds] | 00:37 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 00:47 | |
CIA-18 | shogun: Sergey Lisitsyn master * r4bc9001 / src/configure : Added Eigen3 configure check - http://git.io/--NGKw | 01:21 |
---|---|---|
CIA-18 | shogun: Sergey Lisitsyn master * rc86b318 / src/shogun/io/AsciiFile.cpp : Removed warning-causing empty backslashed line in AsciiFile - http://git.io/UTjdzQ | 01:31 |
CIA-18 | shogun: Sergey Lisitsyn master * r6375108 / (3 files in 2 dirs): Fixed least squares multitask L1/Lq regressor and made it fit with SLEPs analogous method - http://git.io/2mDFcQ | 02:32 |
CIA-18 | shogun: Sergey Lisitsyn master * rf408929 / (3 files in 2 dirs): Merge branch 'slep' of git://github.com/lisitsyn/shogun - http://git.io/g-4awQ | 02:32 |
-!- blackburn [~blackburn@81.28.187.0] has quit [Ping timeout: 246 seconds] | 02:36 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 02:41 | |
CIA-18 | shogun: Sergey Lisitsyn master * r5643f9a / src/shogun/transfer/multitask/TaskGroup.h : Added a few doc lines for task group - http://git.io/yKmFqQ | 02:54 |
CIA-18 | shogun: Sergey Lisitsyn master * rdf424fa / (10 files in 3 dirs): Added tree regularized logistic regression and fixed least squares regressors support - http://git.io/Wk65cw | 04:55 |
CIA-18 | shogun: Sergey Lisitsyn master * r102ecb0 / (6 files in 3 dirs): Added PegasosSVM - http://git.io/8OdO2Q | 06:29 |
CIA-18 | shogun: Sergey Lisitsyn master * rc586669 / (13 files in 6 dirs): Moved liblinear to optimization - http://git.io/LmFqHg | 06:38 |
-!- blackburn [~blackburn@81.28.187.0] has quit [Ping timeout: 250 seconds] | 06:47 | |
shogun-buildbot | build #14 of deb1 - libshogun is complete: Failure [failed compile] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/14 blamelist: Sergey Lisitsyn <lisitsyn.s.o@gmail.com> | 07:10 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 09:32 | |
-!- pluskid [~pluskid@111.120.52.220] has joined #shogun | 09:49 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 13:21 | |
-!- blackburn1 [~blackburn@81.28.187.0] has joined #shogun | 13:30 | |
-!- blackburn1 [~blackburn@81.28.187.0] has quit [Quit: Leaving.] | 13:38 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Ping timeout: 248 seconds] | 13:41 | |
@sonney2k | blackburn, you broke the build... | 13:57 |
@sonney2k | ...instead of fixing it | 13:57 |
blackburn | lol | 13:59 |
blackburn | nahh easy | 13:59 |
@sonney2k | blackburn, please fix it then... we really need to get the buildbot back to green | 14:01 |
blackburn | sonney2k: but we have pegasos now :D | 14:01 |
CIA-18 | shogun: Sergey Lisitsyn master * r88ff286 / (3 files in 3 dirs): Updated headers due to liblinear migration - http://git.io/nEwrMg | 14:03 |
blackburn | should work now | 14:03 |
shogun-buildbot | build #15 of deb1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb1%20-%20libshogun/builds/15 | 14:07 |
CIA-18 | shogun: Sergey Lisitsyn master * r6857283 / (2 files in 2 dirs): Fixed things causing compilation warnings - http://git.io/uCUJCw | 14:10 |
blackburn | sonney2k: hmm heiko stuff fails | 14:10 |
blackburn | the only thing that fails I guess | 14:11 |
blackburn | I disable it for now | 14:14 |
CIA-18 | shogun: Sergey Lisitsyn master * r7ef9212 / (2 files): Disabled statistics tests and updated MT LR test - http://git.io/YIC4mA | 14:15 |
-!- blackburn [~blackburn@81.28.187.0] has left #shogun [] | 14:57 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 16:19 | |
n4nd0 | sonney2k: around? | 16:24 |
-!- pluskid [~pluskid@111.120.52.220] has quit [Quit: Leaving] | 16:46 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 17:29 | |
-!- heiko1 [~heiko@host86-179-33-254.range86-179.btcentralplus.com] has joined #shogun | 19:08 | |
@sonney2k | heiko1, seen what blackburn said? | 19:09 |
@sonney2k | <blackburn> -sonney2k: hmm heiko stuff fails | 19:09 |
@sonney2k | <blackburn> -the only thing that fails I guess | 19:09 |
@sonney2k | <blackburn> -I disable it for now | 19:09 |
@sonney2k | <CIA-18> -shogun: Sergey Lisitsyn master * r7ef9212 / (2 files): Disabled statistics tests and updated MT LR test - http://git.io/YIC4mA | 19:09 |
heiko1 | oh no! | 19:09 |
heiko1 | Ill fix it | 19:10 |
@sonney2k | hmmhh actually sparse stuff is still broken for lots of static interfaces... | 19:11 |
@sonney2k | blackburn - why did you add pegasos? it is the same as SGD but slow | 19:12 |
heiko1 | blackburn, sonney2k, with an updated git version, both statistics examples dont fail here | 19:19 |
heiko1 | what happened? | 19:19 |
@sonney2k | heiko1, did blackburn not just deactivate them in the patch avove? | 19:33 |
heiko1 | yes he did | 19:33 |
heiko1 | I think I know | 19:34 |
heiko1 | they might fail sometimes 1/100 times | 19:35 |
heiko1 | Ill deactivate the assertions | 19:35 |
heiko1 | actually stupid of me leaving them in there | 19:35 |
-!- heiko1 [~heiko@host86-179-33-254.range86-179.btcentralplus.com] has left #shogun [] | 19:36 | |
-!- blackburn [~blackburn@81.28.187.0] has joined #shogun | 19:38 | |
blackburn | argh | 19:38 |
blackburn | sonney2k: I became bored and checked our main page | 19:40 |
blackburn | so we had no pegasos and I thought why not to add it :D | 19:40 |
blackburn | I'll fix sparse don't worry - I just forgot to do it for octave_* | 19:40 |
@sonney2k | blackburn, so you converted it to use dotfeatures? | 19:42 |
blackburn | sonney2k: pegasos? yeah I think so | 19:42 |
@sonney2k | and is it the original algorihtm? | 19:42 |
@sonney2k | I meant what is the source? | 19:43 |
blackburn | a little modified though | 19:43 |
blackburn | from shai shalev | 19:43 |
@sonney2k | hmmhh heiko is gone again | 19:43 |
blackburn | homepage | 19:43 |
@sonney2k | ok | 19:43 |
blackburn | sonney2k: I even didn't really test and compare it | 19:43 |
@sonney2k | blackburn, then file a bug for that | 19:44 |
@sonney2k | otherwise we will forget that this doesn't work :D | 19:44 |
blackburn | sonney2k: I have no idea whether it works | 19:44 |
blackburn | may be it does | 19:44 |
blackburn | :D | 19:44 |
blackburn | sonney2k: insomnia makes some crazy things with me | 19:44 |
blackburn | :D | 19:44 |
@sonney2k | seems so :) | 19:46 |
@sonney2k | better fix the sparse bugs for static/modular interfaces... | 19:46 |
blackburn | sonney2k: will do | 19:47 |
@sonney2k | I would expect there are more in java/ etc to fix | 19:47 |
blackburn | hmm | 19:47 |
@sonney2k | but I guess the bug is always the same | 19:47 |
blackburn | sonney2k: do we have sparse in java? | 19:47 |
@sonney2k | if someone converts dense -> sparse :D | 19:48 |
blackburn | sonney2k: then it is not affected | 19:48 |
blackburn | sonney2k: basically bug is related with MALLOC | 19:48 |
blackburn | sonney2k: we allocate some crap and operator= tries to free it | 19:48 |
blackburn | so what I did is just added inplace new | 19:48 |
@sonney2k | hmmhh wouldn't it be better to have just some simple struct for that internally and for I/O have some overhead with refcounts and copying? | 19:50 |
blackburn | sonney2k: I've never known about inplace new actually and pluskid helped with it | 19:50 |
@sonney2k | I don't like the new[] crap | 19:50 |
blackburn | sonney2k: not new[] | 19:50 |
blackburn | new (&matrix[i]) SGSparseVector<float64_t>(); that is | 19:50 |
blackburn | sonney2k: when you allocate memory for sparse vectors you should initialize it somehow | 19:50 |
blackburn | that thing does this job | 19:51 |
@sonney2k | looks horrible | 19:51 |
@sonney2k | how many people can parse this expression? | 19:51 |
blackburn | sonney2k: there is no other way to have array of SGReferencedData | 19:51 |
@sonney2k | blackburn, that is why I mean we better don't | 19:52 |
blackburn | sonney2k: what do you suggest then? | 19:52 |
@sonney2k | so we could just use some sparse vector struct as we had internally | 19:52 |
blackburn | I'd like to easyfix it now rather than changing internals again | 19:52 |
blackburn | struct like? | 19:52 |
@sonney2k | like we had | 19:53 |
@sonney2k | and then on top of it have the SGReferencedData derived sparse vector class that copies though | 19:53 |
blackburn | I am not so unhappy with it - you only should do it once you allocate memory | 19:53 |
@sonney2k | blackburn, but don't we have the overhead now? | 19:54 |
blackburn | sonney2k: no overhead at all | 19:54 |
@sonney2k | wait - so sparse features uses sparse matrix | 19:54 |
@sonney2k | and sparse matrix doesn't use sparse vector? | 19:54 |
blackburn | sonney2k: sparse features make no use of sparse matrix | 19:54 |
blackburn | only input | 19:55 |
blackburn | it contains array of sparse vectors | 19:55 |
blackburn | so each time you allocate new array you have to initialize refs | 19:55 |
@sonney2k | blackburn, well it should make use of sparse matrix | 19:55 |
blackburn | sonney2k: it will be the same - but in sparse matrix then | 19:56 |
@sonney2k | like CDenseFeatures | 19:56 |
blackburn | sonney2k: you *have* to new each sparse vector of array | 19:57 |
blackburn | because operator= will corrupt it in other case | 19:57 |
@sonney2k | blackburn, well you don't if we don't use SGReferecedData right? | 19:57 |
blackburn | yes | 19:58 |
@sonney2k | so now we add the overhead to each sparse vector | 19:58 |
blackburn | ??? | 19:58 |
blackburn | overhead like? | 19:58 |
blackburn | reference count? | 19:58 |
@sonney2k | yes | 19:59 |
blackburn | sonney2k: I thought we discissed it is ok | 19:59 |
@sonney2k | I don't remember... | 20:00 |
@sonney2k | it is 12 byte / vector | 20:00 |
@sonney2k | lets consider 100 Mio examples | 20:01 |
@sonney2k | so just 1.2G | 20:01 |
@sonney2k | looks OK I gues | 20:01 |
@sonney2k | ss | 20:01 |
blackburn | sonney2k: but much easier to work with | 20:02 |
@sonney2k | blackburn, yeah except for expressions like (&sparse_matrix[i])->~SGSparseVector<T>(); and the one above :/ | 20:05 |
blackburn | sonney2k: hah well I can live with it - it appears only once | 20:05 |
blackburn | we have some uglier code | 20:05 |
@sonney2k | blackburn, where ? | 20:06 |
@sonney2k | well ok if it just appears once | 20:07 |
@sonney2k | blackburn, so then convert the sparsefeatures also to use sgsparsematrix | 20:07 |
@sonney2k | then it is ok | 20:07 |
blackburn | sonney2k: pthreads are unparseable | 20:07 |
@sonney2k | yeah but that is easily fixable | 20:07 |
@sonney2k | if we treat sparse features like this | 20:07 |
@sonney2k | we should do the same with strings later | 20:08 |
blackburn | yeah but I'd like to not convert sparse features just now | 20:08 |
@sonney2k | blackburn, btw I really wonder if we shouldn't have something like that instead of the dense feature matrix | 20:09 |
CIA-18 | shogun: Sergey Lisitsyn master * r6637f20 / (2 files in 2 dirs): Fixed sparse matrix handling in octave interfaces - http://git.io/JpzEsw | 20:09 |
@sonney2k | I mean also just list of vectors | 20:09 |
blackburn | why? | 20:09 |
@sonney2k | one can permute vectors nicely... | 20:09 |
blackburn | well it makes sense for me actually | 20:10 |
@sonney2k | on the other hand blas/lapack stuff needs such fortran style matrice | 20:10 |
@sonney2k | s | 20:10 |
blackburn | sonney2k: well we rarely work on full matrices | 20:20 |
blackburn | sonney2k: do we work on matrices at all? | 20:20 |
@sonney2k | blackburn, don't GMM / GPs do? | 20:23 |
@sonney2k | blackburn, I guess not an issue now though | 20:23 |
@sonney2k | much more important is get things back to stable | 20:24 |
@sonney2k | then sparsefeatures -> sparsematrix, GMMs, buildbots, tests running lalal | 20:24 |
@sonney2k | a | 20:24 |
blackburn | sonney2k: lets see if tests will go now | 20:25 |
@sonney2k | examples you mean :) | 20:25 |
@sonney2k | tests are all disabled unfortunately :/ | 20:26 |
blackburn | nearly no difference | 20:26 |
@sonney2k | hmmhh | 20:27 |
@sonney2k | I would wish | 20:27 |
@sonney2k | last time we said that too :D | 20:27 |
@sonney2k | and we were fixing things for a few weeks | 20:27 |
blackburn | if it compiles - nice | 20:27 |
blackburn | :D | 20:27 |
blackburn | if it does not produce segfaults - awesome :D | 20:28 |
blackburn | sonney2k: what is treesize? | 20:30 |
blackburn | 573276 KiB?? | 20:30 |
blackburn | 573 mb? | 20:30 |
@sonney2k | yes | 20:31 |
@sonney2k | the compiled stuff including data | 20:31 |
blackburn | pretty much | 20:32 |
blackburn | argh | 20:47 |
blackburn | that didn't help | 20:47 |
blackburn | beautiful | 21:00 |
blackburn | sonney2k: any idea what can be wrong with kernel_lik? | 21:23 |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has joined #shogun | 21:44 | |
-!- puffin444 [62e3926e@gateway/web/freenode/ip.98.227.146.110] has quit [Quit: Page closed] | 23:01 | |
CIA-18 | shogun: Sergey Lisitsyn master * rf482b5d / examples/undocumented/matlab_and_octave/kernel_lik.m : Temporary commented out kernel_lik.m example - http://git.io/hNnE9w | 23:20 |
--- Log closed Mon Jul 02 00:00:17 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!