--- Log opened Sat Oct 25 00:00:39 2014 | ||
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has joined #shogun | 00:40 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 00:40 | |
@HeikoS | wiking: ping? | 00:55 |
---|---|---|
-!- iglesiasg [~iglesias@50-76-45-78-ip-static.hfc.comcastbusiness.net] has joined #shogun | 00:55 | |
@HeikoS | iglesiasg: hello :) | 00:55 |
-!- jiaolong [5c15c2a6@gateway/web/freenode/ip.92.21.194.166] has joined #shogun | 01:35 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has quit [Quit: Leaving.] | 01:49 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has joined #shogun | 02:00 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 02:00 | |
-!- jiaolong [5c15c2a6@gateway/web/freenode/ip.92.21.194.166] has quit [Quit: Page closed] | 02:11 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has quit [Ping timeout: 265 seconds] | 02:17 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has joined #shogun | 02:21 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 02:21 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has quit [Ping timeout: 265 seconds] | 02:33 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has joined #shogun | 02:38 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 02:38 | |
-!- iglesiasg [~iglesias@50-76-45-78-ip-static.hfc.comcastbusiness.net] has quit [Quit: leaving] | 02:38 | |
-!- HeikoS [~heiko@50-76-45-78-ip-static.hfc.comcastbusiness.net] has quit [Ping timeout: 255 seconds] | 02:43 | |
shogun-buildbot | build #907 of nightly_default is complete: Failure [failed notebooks] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/907 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Fernando Iglesias <fernando.iglesiasg@gmail.com>, Wu Lin <yorker.lin@gmail.com>, tklein23 <tklein23@users.noreply.github.com>, Thoralf Klein | 04:35 |
shogun-buildbot | <thoralf@fischlustig.de> | 04:35 |
-!- HeikoS [~heiko@107.17.164.143] has joined #shogun | 08:26 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 08:26 | |
-!- iglesiasg [~iglesias@12.104.145.5] has joined #shogun | 08:37 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 08:37 | |
@iglesiasg | Gentlemen | 08:37 |
@HeikoS | iglesiasg: good evening | 08:42 |
@wiking | :) | 08:49 |
@HeikoS | wiking: heya | 08:49 |
@wiking | where r u guys? :) | 08:49 |
@HeikoS | wiking: san jose | 08:49 |
@HeikoS | iglesiasg, https://github.com/shogun-toolbox/shogun/wiki/SWIG-issues | 08:49 |
@wiking | ah ok thatswhy it's good evening | 08:49 |
@wiking | it's 3pm here | 08:49 |
@wiking | HeikoS: for swig issues we shoulg go with d-poiinters | 08:51 |
@HeikoS | wiking: 3pm nice | 08:51 |
@HeikoS | wiking: yeah we thought about that too | 08:51 |
@HeikoS | wiking: what about writing some things down | 08:51 |
@HeikoS | wiking: so that its not just all in our heads | 08:51 |
@HeikoS | and maybe we can make some entrance tasks from this | 08:53 |
@HeikoS | wiking: could you explain the d-pointer stuff a bit? | 08:53 |
@wiking | mmm not now | 08:53 |
@wiking | still working here :) | 08:53 |
@wiking | but here's a good article about it | 08:54 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 08:54 | |
shogun-notifier- | shogun: jiaolong :develop * 05ce392 / src/shogun/structure/GraphCut.h: https://github.com/shogun-toolbox/shogun/commit/05ce392abb1871c8c73c217914812ee88562675d | 08:54 |
shogun-notifier- | shogun: removed buildbot warnings from GraphCut.h | 08:54 |
shogun-notifier- | shogun: Heiko Strathmann :develop * ef2a460 / src/shogun/structure/GraphCut.h: https://github.com/shogun-toolbox/shogun/commit/ef2a460b6121ce358266f2cca71ce4d443592c6b | 08:54 |
shogun-notifier- | shogun: Merge pull request #2554 from Jiaolong/rm_warnings | 08:54 |
shogun-notifier- | shogun: | 08:54 |
shogun-notifier- | shogun: removed buildbot warnings from GraphCut.h | 08:54 |
@wiking | http://qt-project.org/wiki/Dpointer | 08:54 |
@wiking | http://zchydem.enume.net/2010/01/19/qt-howto-private-classes-and-d-pointers/ | 08:54 |
@HeikoS | wiking: thanks | 08:54 |
shogun-buildbot | build #81 of trusty - libshogun - viennacl is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/81 | 08:58 |
shogun-buildbot | build #845 of precise - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/845 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, jiaolong <jiaolongxu@gmail.com> | 09:01 |
-!- travis-ci [~travis-ci@ec2-23-22-26-147.compute-1.amazonaws.com] has joined #shogun | 09:06 | |
travis-ci | it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/38997367 | 09:06 |
-!- travis-ci [~travis-ci@ec2-23-22-26-147.compute-1.amazonaws.com] has left #shogun [] | 09:06 | |
shogun-buildbot | build #873 of FCRH - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/873 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, jiaolong <jiaolongxu@gmail.com> | 09:20 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 096de32 / src/shogun/preprocessor/PCA.h: https://github.com/shogun-toolbox/shogun/commit/096de322b6f23d231a9db5408698cf78c44bda81 | 10:13 |
shogun-notifier- | shogun: little doc correction | 10:13 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 3f26225 / src/shogun/mathematics/linalg/dotproduct/VectorDotProduct.h: https://github.com/shogun-toolbox/shogun/commit/3f262258baf70ecb03f241685c9c5dc23124544b | 10:13 |
shogun-notifier- | shogun: fixed doxygen math | 10:13 |
shogun-notifier- | shogun: Heiko Strathmann :develop * cd805c7 / src/shogun/clustering/ (2 files): https://github.com/shogun-toolbox/shogun/commit/cd805c75f0b6266dde13d19df4995026a62ab0ab | 10:13 |
shogun-notifier- | shogun: removed stdio from header and add documentation string | 10:13 |
shogun-notifier- | shogun: Heiko Strathmann :develop * cb4672f / src/shogun/clustering/KMeans.h: https://github.com/shogun-toolbox/shogun/commit/cb4672f442556425d20edcd99f385ec11e019a7a | 10:13 |
shogun-notifier- | shogun: added some mini updates to doc | 10:13 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 907b8da / src/shogun/distributions/KernelDensity.h: https://github.com/shogun-toolbox/shogun/commit/907b8da70ad2267d0b35906214cec0c9d93b295f | 10:13 |
shogun-notifier- | shogun: doc string | 10:13 |
shogun-notifier- | shogun: Heiko Strathmann :develop * d1384f9 / src/shogun/structure/libbmrm.h: https://github.com/shogun-toolbox/shogun/commit/d1384f9b7e19fb106480bb96cbfadc91e60b842e | 10:13 |
shogun-notifier- | shogun: doc string | 10:13 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 24bbcc5 / src/shogun/distributions/EMBase.h: https://github.com/shogun-toolbox/shogun/commit/24bbcc52b7ff73c7b2a5d6e1bd4d146f38da6e85 | 10:13 |
shogun-notifier- | shogun: doc strings | 10:14 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 104a258 / src/shogun/distributions/EMBase.h,src/shogun/distributions/EMMixtureModel.h: https://github.com/shogun-toolbox/shogun/commit/104a25851abc69bbf2be457c0bfc9847a0851078 | 10:14 |
shogun-notifier- | shogun: docstrings | 10:14 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 90d1a6f / src/shogun/labels/MultilabelLabels.h: https://github.com/shogun-toolbox/shogun/commit/90d1a6f5cb5a0523beda10bd21cf774b47d75ee4 | 10:14 |
shogun-notifier- | shogun: docstrings | 10:14 |
-!- witness___ [uid10044@gateway/web/irccloud.com/x-rbwdfoufhxjylbfe] has quit [Quit: Connection closed for inactivity] | 10:20 | |
shogun-buildbot | build #78 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/78 blamelist: jiaolong <jiaolongxu@gmail.com> | 10:28 |
-!- Phawx [~Phawx@178.19.210.162] has joined #shogun | 10:32 | |
shogun-buildbot | build #79 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/79 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 10:40 |
-!- HeikoS [~heiko@107.17.164.143] has quit [Quit: Leaving.] | 11:01 | |
-!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has joined #shogun | 11:13 | |
-!- Phawx [~Phawx@178.19.210.162] has quit [Quit: Leaving] | 11:39 | |
@lisitsyn | iglesiasg: you and heiko should answer my super question ;) | 12:02 |
@lisitsyn | think about it! | 12:02 |
@lisitsyn | what's next after deep learning and how can we jump on this train | 12:02 |
@wiking | hahaha | 12:08 |
-!- travis-ci [~travis-ci@ec2-23-22-26-147.compute-1.amazonaws.com] has joined #shogun | 12:50 | |
travis-ci | it's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: http://travis-ci.org/shogun-toolbox/shogun/builds/38999536 | 12:50 |
-!- travis-ci [~travis-ci@ec2-23-22-26-147.compute-1.amazonaws.com] has left #shogun [] | 12:50 | |
shogun-buildbot | build #443 of debian wheezy - memcheck is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/443 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, jiaolong <jiaolongxu@gmail.com> | 12:57 |
shogun-buildbot | build #846 of precise - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/846 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 13:02 |
shogun-buildbot | build #847 of precise - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/847 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 13:02 |
shogun-buildbot | build #874 of FCRH - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/874 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 13:06 |
@iglesiasg | shit, I fell asleep while compiling | 13:08 |
@wiking | lol | 13:08 |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 13:14 | |
shogun-buildbot | build #80 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/80 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 13:34 |
shogun-buildbot | build #81 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/81 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 13:59 |
@iglesiasg | when we build PythonModular with Eigen support we get some Eigen__MatrixXt stuff in modshogunPYTHON_wrap.cxx | 14:06 |
@iglesiasg | this is not there when building w/o Eigen | 14:07 |
@wiking | iglesiasg: mmm then that means that eigen is being imported in a header | 14:11 |
@wiking | somewhere... | 14:11 |
@iglesiasg | wiking: I checked that with grep and didn't find it, so at least no direct include | 14:12 |
@iglesiasg | grep -R eigen */*.h -- that's how I checked it | 14:12 |
@wiking | git grep eigen -- '*.h' | 14:13 |
@iglesiasg | -Ri actually, anyway | 14:13 |
@wiking | in src/shogun | 14:13 |
@wiking | and you'll see quite some | 14:13 |
@wiking | git grep eigen3.h -- '*.h | 14:13 |
@wiking | machine/gp/SingleLaplacianInferenceMethodWithLBFGS.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/dotproduct/DenseEigen3DotProduct.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/Add.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@iglesiasg | yep | 14:14 |
@wiking | mathematics/linalg/internal/implementation/Dot.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/ElementwiseProduct.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/ElementwiseSquare.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/MatrixProduct.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/Max.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/Scale.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/SetRowsConst.h:#include <shogun/mathematics/eigen3.h> | 14:14 |
@wiking | mathematics/linalg/internal/implementation/SpecialPurpose.h:#include <shogun/mathemat | 14:14 |
@wiking | the linalg shouldn't matter | 14:14 |
@wiking | at least not the internal one | 14:14 |
@wiking | i reckon you can throw it out from machine/gp/SingleLaplacianInferenceMethodWithLBFGS.h | 14:14 |
@iglesiasg | let me try that | 14:15 |
@wiking | it's a common mistake done by Wu Lin | 14:15 |
@wiking | he always did this regardless how many times i told not to do this | 14:15 |
@wiking | ah now | 14:15 |
@wiking | *no | 14:16 |
@wiking | he fucked it up | 14:16 |
@wiking | /* compute the gradient given the current alpha*/ | 14:16 |
@wiking | void get_gradient_wrt_alpha(Eigen::Map<Eigen::VectorXd>* alpha, | 14:16 |
@wiking | Eigen::Map<Eigen::VectorXd>* gradient); | 14:16 |
@wiking | /* compute the function value given the current alpha*/ | 14:16 |
@wiking | void get_psi_wrt_alpha(Eigen::Map<Eigen::VectorXd>* alpha, | 14:16 |
@wiking | float64_t* psi); | 14:16 |
@iglesiasg | arrrrgh | 14:16 |
@wiking | this is a private function in the header | 14:16 |
@iglesiasg | that is not nice in the header | 14:16 |
@wiking | but i reckon you can hide this | 14:16 |
@wiking | mmm | 14:17 |
@wiking | well i worst case what can be done | 14:17 |
@wiking | is to make this into a static function | 14:17 |
@wiking | where u pass all the variables of the class | 14:17 |
@wiking | get_psi_wrt_alpha uses some of the object variables | 14:17 |
@wiking | same goes for get_gradient_wrt_alpha | 14:18 |
@wiking | ok now | 14:18 |
@wiking | this should be fairly easy to fix | 14:19 |
@wiking | Eigen::Map<Eigen::VectorXd> eigen_alpha(alpha_cast, dim); | 14:19 |
@wiking | float64_t psi = 0.0; | 14:19 |
@wiking | obj_prt->get_psi_wrt_alpha(&eigen_alpha, &psi); | 14:19 |
@wiking | Eigen::Map<Eigen::VectorXd> eigen_gradient(gradient, dim); | 14:19 |
@wiking | obj_prt->get_gradient_wrt_alpha(&eigen_alpha, &eigen_gradient); | 14:19 |
@wiking | as these are all maps of some very standard arrays | 14:19 |
@wiking | so just do the mapping within those functions :P | 14:19 |
lambday | wiking: iglesiasg: I just removed this one - mathematics/linalg/dotproduct/DenseEigen3DotProduct.h | 14:20 |
lambday | this one is deprecated and is no longer needed | 14:20 |
@wiking | lambday: commited? | 14:20 |
@wiking | if so then just push it | 14:20 |
lambday | wiking: didn't push yet - still fixing some doxygen bugs | 14:20 |
lambday | alright | 14:20 |
lambday | pushing whatever I have till now | 14:20 |
@wiking | and once iglesiasg fixes that then swig wont generate eigen mappings :P | 14:20 |
@iglesiasg | wiking: why the linalg ones shouldn't matter though? They're just skipped by swig, I am guessing | 14:21 |
@wiking | iglesiasg: because those headers are used internally in the linalg part of shgun | 14:21 |
@wiking | i.e. we don't do mappings over that | 14:21 |
@wiking | iglesiasg: so it's almost like d-pointers | 14:21 |
@iglesiasg | allright | 14:22 |
@wiking | lambday: how about mathematics/linalg/linsolver/IterativeSolverIterator.h | 14:22 |
@wiking | ? | 14:22 |
lambday | wiking: that one includes eigen3 in the header? | 14:23 |
@wiking | but i reckon we don't do mappings of that either or? | 14:23 |
lambday | wiking: let me check | 14:23 |
@wiking | mathematics/linalg/linsolver/IterativeSolverIterator.h:#include <shogun/mathematics/eigen3.h> | 14:23 |
@wiking | no we are good | 14:23 |
lambday | wiking: should be avoided... when i did that I didn't know much | 14:23 |
lambday | will fix | 14:23 |
@wiking | cool | 14:24 |
@wiking | iglesiasg: although i wonder about SingleLaplacianInferenceMethodWithLBFGS as well.... | 14:24 |
lambday | wiking: but iirc then iterative solver iterator is used internally | 14:24 |
lambday | I mean, its not exposed | 14:24 |
@wiking | oh nothing... | 14:24 |
@wiking | iglesiasg: | 14:24 |
@wiking | interfaces/modular/GaussianProcess.i:%rename(SingleLaplacianInferenceMethodWithLBFGS) CSingleLaplacianInferenceMethodWithLBFGS; | 14:24 |
@wiking | interfaces/modular/GaussianProcess.i:%include <shogun/machine/gp/SingleLaplacianInferenceMethodWithLBFGS.h> | 14:24 |
@wiking | interfaces/modular/GaussianProcess_includes.i: #include <shogun/machine/gp/SingleLaplacianInferenceMethodWithLBFGS.h> | 14:24 |
@wiking | this is why eigen is actually in the modshogun_PYTHON.cxx | 14:25 |
@wiking | git grep IterativeSolverIterator -- '*.i' | 14:25 |
@wiking | is an empty set | 14:25 |
lambday | wiking: yeah that's what I thought - shouldn't matter | 14:25 |
@wiking | indeed | 14:26 |
@wiking | so maybe once iglesiasg fixes that shit in SingleLaplacianInferenceMethodWithLBFGS we'll have travis working again :) | 14:26 |
@wiking | \o/ | 14:26 |
lambday | wiking: btw about khaled's random matrix/vector patch | 14:31 |
lambday | wiking: as of now I think its better to keep it within linalg | 14:32 |
lambday | wiking: Heiko suggested something - that we maybe want to put it alongside Random.h | 14:32 |
lambday | but this thing heavily uses viennacl and then we'll have viennacl code in many places apart from linalg | 14:33 |
lambday | wiking: btw I just noticed in some of the linalg functions khaled has kept generic implementation as Eigen3 specialization... maybe I'll change those | 14:33 |
lambday | SpecialPurpose.h.. Core.h etc | 14:34 |
lambday | iglesiasg: wiking: for eigen mapping stuff, we can use SGMatrix's map operator and Eigen constructor.. its neater | 14:40 |
@iglesiasg | lambday: I am following the style how Wu did it in other parts of that file and just pass the float* and the dimension around | 14:41 |
@iglesiasg | lambday: but I agree with you, it's better that other way | 14:42 |
lambday | iglesiasg: yeah maybe for next tasks we can. | 14:43 |
@iglesiasg | in any case, this code has really weird things... | 14:44 |
@iglesiasg | I don't really see the point of passing around Eigen::VectorXd* alpha and do the math operations as *alpha | 14:44 |
lambday | iglesiasg: I guess eigen3 does deep copy.. maybe use reference? | 14:50 |
@iglesiasg | lambday: I'd guess that it doesn't do deep copy, but I am not sure | 14:51 |
lambday | iglesiasg: let me check | 14:51 |
@wiking | map is not a deep copy | 14:51 |
@wiking | just a wrapping | 14:51 |
@iglesiasg | lambday: anyhow, since I am passing the float* together with the dimension, I don't need to be aware of that just yet | 14:51 |
lambday | iglesiasg: I remember khaled saying this once | 14:51 |
lambday | nah map is not | 14:52 |
@iglesiasg | is not deep copy? | 14:52 |
lambday | map is not deep copy.. | 14:52 |
@wiking | Eigen::Map<Eigen::VectorXd> | 14:52 |
@iglesiasg | right | 14:52 |
@wiking | it's a simple wrapping | 14:52 |
@wiking | nothing else | 14:52 |
lambday | but VectorXd vec1 = vec2 does deep copy | 14:52 |
@iglesiasg | I hope this SingleLaplacianInferenceMethodWithLBFGS is unit tested | 14:53 |
@iglesiasg | otherwise this refactoring might end up biting me | 14:53 |
@iglesiasg | yeah, it seems to have proper tests, awesome | 14:56 |
lambday | iglesiasg: it does deep copy when you pass Eigen::Matrix as param.. just checked | 14:56 |
@iglesiasg | lambday: will try to keep that in mind from now on, thanks! | 14:57 |
lambday | iglesiasg: I don't understand why should one pass *float with the dimension? :( | 14:58 |
@iglesiasg | btw, they brought Linus and Peter Norvig yesterday | 15:00 |
@iglesiasg | lambday: yeah.. I agree with you. It's not the proper way of doing it | 15:00 |
@iglesiasg | lambday: what does rows() return for an Eigen::Map<Eigen::VectorXd>? | 15:03 |
@iglesiasg | lambday: just the vector dimension, right? | 15:04 |
lambday | iglesiasg: yes | 15:04 |
@iglesiasg | cool, thanks | 15:04 |
lambday | iglesiasg: I guess the easiest way would be to change the function params to SG.. even if you pass Eigen matrix due to the eigen constructor in SG it should do the conversion automatically... then inside the function we can use map from SG | 15:05 |
lambday | oh but it uses eigen map instead.. that won't work then | 15:07 |
lambday | sorry about the confusion | 15:07 |
@wiking | i'm totally puzzled by this issues | 15:09 |
@wiking | https://github.com/shogun-toolbox/shogun/issues/2538 | 15:09 |
@wiking | iglesiasg: still here? | 15:13 |
@iglesiasg | yep | 15:13 |
@iglesiasg | the Eigen__MatrixXt thingies still appear in the modshogunPYTHON* | 15:14 |
@iglesiasg | I am double checking it though.. | 15:15 |
@wiking | iglesiasg: do you know why SGString doesn't have a destructor? | 15:15 |
@iglesiasg | amazing | 15:15 |
@iglesiasg | wiking: destroy_string() xD | 15:16 |
@wiking | iglesiasg: ok yeah but like a real dtor | 15:16 |
@wiking | ~SGString() | 15:16 |
@wiking | ok i'll make one :) | 15:17 |
@iglesiasg | wiking: yeah, I got you. I found it funny that there's this destroy_string that is called from nowhere haha | 15:17 |
@wiking | iglesiasg: mmm maybe i should just drop then destroy_string and free_string | 15:18 |
@wiking | mmm ok no | 15:18 |
@wiking | i'll just remove destroy_string | 15:18 |
@wiking | lambday: how do i force to turn off viennacl? | 15:30 |
@wiking | -DVIENNACL_FOUND=OFF doesn't work | 15:30 |
lambday | wiking: checking | 15:31 |
lambday | wiking: HAVE_VIENNACL is the macro... | 15:32 |
lambday | wiking: I'm afraid current it's not done | 15:34 |
lambday | wiking: if the package is not found only then it will say viennacl missing | 15:34 |
lambday | CMakeLists.txt:651 | 15:35 |
lambday | wiking: maybe define another variable for this? if that's turned off then we'll find the viennacl pkg | 15:36 |
@iglesiasg | wiking: so it was not only this thing in SingleLaplacianInferenceMethodWithLBFGS | 15:42 |
@iglesiasg | eigen is leaking into swig somewhere else too | 15:43 |
@wiking | :< | 15:43 |
@wiking | fuck | 15:43 |
@wiking | iglesiasg: have u compiled the thing from scratch | 15:43 |
@wiking | rm -rf build | 15:43 |
@iglesiasg | rm -rf build | 15:43 |
@iglesiasg | yes | 15:43 |
@iglesiasg | but it is so nuts this thing of the memory comsumption w/ and w/o eigen | 15:44 |
@iglesiasg | 2GB and ~0.5GB | 15:44 |
@wiking | fuck | 15:44 |
@wiking | mmm is it really 0.5GB | 15:44 |
@wiking | i never seen swig requiring oonly 0.5 :P | 15:44 |
@iglesiasg | wiking: compile PythonModular w/o Eigen :-) | 15:45 |
@lisitsyn | hah | 15:45 |
@lisitsyn | really? | 15:45 |
@iglesiasg | I wanted to double check this though in case I missed something yesterday | 15:45 |
@lisitsyn | swig + eigen = 2gb of ram? | 15:45 |
@iglesiasg | let me do it now | 15:45 |
@lisitsyn | and w/o just 0.5? | 15:45 |
@iglesiasg | lisitsyn: let me double check, I might be saying just bullshit | 15:46 |
@lisitsyn | iglesiasg: I wish it is true as it sounds awesome | 15:46 |
@lisitsyn | iglesiasg: what if we remove std? :D | 15:46 |
@lisitsyn | 0.1gb? ;) | 15:46 |
@wiking | i mean the thing is that afaik modular interface packages are not being generated for ubuntu | 15:46 |
@wiking | because it uses 4G+ ram | 15:46 |
shogun-buildbot | build #444 of debian wheezy - memcheck is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/444 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 16:21 |
@iglesiasg | wiking: lisitsyn: mistery solved | 16:35 |
@iglesiasg | it was ccache.. | 16:35 |
@wiking | iglesiasg: lol | 16:37 |
@wiking | iglesiasg: ok push the fix then | 16:37 |
@wiking | let's see how travis reacts | 16:37 |
@iglesiasg | wiking: all right | 16:37 |
@wiking | as usually the modular jobs were dying because of insufficient mem | 16:37 |
@iglesiasg | wiking: not very strict measure probably, but it reduced it about 100mb | 16:38 |
@iglesiasg | I mean the refactoring | 16:38 |
@iglesiasg | wiking: but I think it is worth to investigate this thing of Eigen__MatrixXt in modshogunPYTHON_wrap.cxx. I don't see any reason why there should any Eigen related thing in there | 16:39 |
@iglesiasg | wiking: what do you think? | 16:39 |
@wiking | mmm | 16:40 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 16:40 | |
shogun-notifier- | shogun: lambday :develop * f918c77 / src/shogun/mathematics/linalg/dotproduct/ (4 files): https://github.com/shogun-toolbox/shogun/commit/f918c774359cde3fa811ba3f31ef83d68de6b1ac | 16:40 |
shogun-notifier- | shogun: Removed deprecated dot-product from linalg | 16:40 |
shogun-notifier- | shogun: lambday :develop * e223ec7 / src/shogun/lib/ (4 files): https://github.com/shogun-toolbox/shogun/commit/e223ec7e374e2f8468d334c339d2eeb732cac61e | 16:40 |
shogun-notifier- | shogun: minor documentation fix in SG/GPU-Matrix/Vector classes | 16:40 |
shogun-notifier- | shogun: lambday :develop * b73e2f0 / src/shogun/machine/BaggingMachine.h: https://github.com/shogun-toolbox/shogun/commit/b73e2f093fd3df2dbb66537e7320c012c4abde5b | 16:40 |
shogun-notifier- | shogun: removed doxygen warnings | 16:40 |
shogun-notifier- | shogun: lambday :develop * f56294f / src/shogun/mathematics/linalg/ratapprox/logdet/LogDetEstimator.h: https://github.com/shogun-toolbox/shogun/commit/f56294f88af504f48b147d4c0105ccf518d952ee | 16:40 |
shogun-notifier- | shogun: removed doxygen warnings from log-det estimator | 16:40 |
shogun-notifier- | shogun: Soumyajit De :develop * d82150a / src/shogun/ (10 files): https://github.com/shogun-toolbox/shogun/commit/d82150a9d69f5dd04f415d9064d273012cf19118 | 16:40 |
shogun-notifier- | shogun: Merge pull request #2561 from lambday/develop | 16:40 |
shogun-notifier- | shogun: | 16:40 |
shogun-notifier- | shogun: Minor doxy fixes+removed deprecated DotProduct impl from linalg | 16:40 |
lambday | iglesiasg: what did you configure shogun with? I cannot see Eigen__MatrixXt in my modshogunPYTHON_wrap.cxx | 16:43 |
@iglesiasg | lambday: cmake -DPythonModular=ON -DCMAKE_BUILD_TYPE=Debug -DBUNDLE_EIGEN=ON -DENABLE_TESTING=OFF -DBUILD_EXAMPLES=OFF .. | 16:45 |
@iglesiasg | lambday: there was a type, sorry | 16:46 |
@iglesiasg | lambday: Eigen__MatrixT | 16:46 |
@iglesiasg | typo* | 16:46 |
@iglesiasg | saying there's a typo with a typo \o/ | 16:46 |
lambday | iglesiasg: haha :D | 16:47 |
-!- travis-ci [~travis-ci@ec2-54-80-122-236.compute-1.amazonaws.com] has joined #shogun | 16:50 | |
travis-ci | it's Soumyajit De'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/39014672 | 16:50 |
-!- travis-ci [~travis-ci@ec2-54-80-122-236.compute-1.amazonaws.com] has left #shogun [] | 16:50 | |
@iglesiasg | lambday: you got that one in modshogunPYTHON_wrap, right? | 16:50 |
lambday | iglesiasg: yep... | 16:50 |
@iglesiasg | btw, it is king of annoying that enable_testing=on build c++ examples even if build_examples=off | 16:52 |
@iglesiasg | lol king of annoying, I am pushing my boundaries today | 16:52 |
lambday | haha.. you're right actually... its the king of all annoying things out there :D | 16:56 |
@iglesiasg | wiking: do you think we can get a quick fix for that? pleaseee :) | 16:56 |
shogun-buildbot | build #848 of precise - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/848 blamelist: lambday <heavensdevil6909@gmail.com> | 16:59 |
shogun-buildbot | build #849 of precise - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/849 blamelist: lambday <heavensdevil6909@gmail.com>, Soumyajit De <heavensdevil6909@gmail.com> | 17:02 |
lambday | iglesiasg: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/lib/GPUMatrix.h#L115 | 17:06 |
lambday | iglesiasg: this is not SWIG guarded | 17:06 |
lambday | iglesiasg: and here https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/lib/GPUVector.h#L110 | 17:07 |
lambday | iglesiasg: could you please try ifndefing it once? | 17:08 |
@iglesiasg | lambday: yes, I can do that | 17:11 |
lambday | iglesiasg: these might be the culprit here.. | 17:12 |
@iglesiasg | lambday: we were discussing about this kind of thing yesterday, this is not really the best way of handling this issue, right? | 17:12 |
@iglesiasg | lambday: I mean, these classes that need eigen or other dependencies should just not be part of an interface | 17:13 |
@iglesiasg | no matter whether this interface is going to be visible for swig or for users | 17:13 |
@iglesiasg | lambday: what do you think? | 17:13 |
lambday | iglesiasg: well, this stuff should really not be exposed to swig - its used internally in linalg and which makes it actually possible to work with different types of matrices in Eigen and ViennaCL | 17:13 |
@iglesiasg | lambday: exactly | 17:14 |
@iglesiasg | lambday: dumb question coming.. why is it exposed to swig, then? :) | 17:14 |
shogun-buildbot | build #62 of CentOS7 - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/CentOS7%20-%20libshogun/builds/62 blamelist: lambday <heavensdevil6909@gmail.com> | 17:15 |
shogun-buildbot | build #875 of FCRH - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/875 blamelist: lambday <heavensdevil6909@gmail.com> | 17:15 |
lambday | iglesiasg: my bad... when khaled sent this PR I didn't notice this.. | 17:16 |
lambday | iglesiasg: he handled it correctly for SGMatrix and SGVector | 17:16 |
lambday | which has similar things | 17:16 |
lambday | via #ifndef SWIG | 17:16 |
lambday | iglesiasg: gotta go for now.. brb after dinner | 17:17 |
shogun-buildbot | build #63 of CentOS7 - libshogun is complete: Success [build successful] Build details are at http://buildbot.shogun-toolbox.org/builders/CentOS7%20-%20libshogun/builds/63 | 17:20 |
shogun-buildbot | build #876 of FCRH - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/876 blamelist: lambday <heavensdevil6909@gmail.com>, Soumyajit De <heavensdevil6909@gmail.com> | 17:20 |
shogun-notifier- | shogun: Fernando Iglesias :develop * 82202b7 / src/shogun/machine/gp/ (2 files): https://github.com/shogun-toolbox/shogun/commit/82202b75c5cda105eec83a9f518bcf1d892227a0 | 17:20 |
shogun-notifier- | shogun: Refactor part of Laplacian LBFGS to remove eigen3 include in header. | 17:20 |
@iglesiasg | lambday: enjoy your meal | 17:20 |
@iglesiasg | lambday: but then we can just remove these classes from the .i files if they shouldn't be exposed in interfaces | 17:21 |
@iglesiasg | wiking: patch is in, let's see what travis thinks | 17:21 |
@iglesiasg | lambday: they are indeed not in the .i files. Still, it might be the cause of Eigen__MatrixT in swig if they are included in another header that swig has to look at. I will test it and get back later | 17:26 |
-!- travis-ci [~travis-ci@ec2-54-162-62-69.compute-1.amazonaws.com] has joined #shogun | 17:36 | |
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/39016941 | 17:36 |
-!- travis-ci [~travis-ci@ec2-54-162-62-69.compute-1.amazonaws.com] has left #shogun [] | 17:36 | |
@iglesiasg | wiking: no luck yet | 17:40 |
shogun-buildbot | build #82 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/82 blamelist: lambday <heavensdevil6909@gmail.com> | 18:27 |
-!- iglesiasg [~iglesias@12.104.145.5] has quit [Ping timeout: 240 seconds] | 18:39 | |
shogun-buildbot | build #83 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/83 blamelist: lambday <heavensdevil6909@gmail.com>, Soumyajit De <heavensdevil6909@gmail.com> | 18:50 |
shogun-buildbot | build #850 of precise - libshogun is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/precise%20-%20libshogun/builds/850 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 19:13 |
shogun-buildbot | build #877 of FCRH - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/FCRH%20-%20libshogun/builds/877 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 19:14 |
-!- iglesiasg [~iglesias@12.104.145.5] has joined #shogun | 19:16 | |
shogun-buildbot | build #84 of osx2 - modular_interfaces is complete: Failure [failed csharp modular] Build details are at http://buildbot.shogun-toolbox.org/builders/osx2%20-%20modular_interfaces/builds/84 blamelist: Fernando Iglesias <fernando.iglesiasg@gmail.com> | 19:53 |
-!- HeikoS [~heiko@12.104.145.5] has joined #shogun | 19:58 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 19:58 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 20:20 | |
@lisitsyn | HeikoS: iglesiasg: https://github.com/shogun-toolbox/shogun/issues/2564 | 21:29 |
iglesiasg | lisitsyn: why would it compile swig faster? | 21:39 |
shogun-buildbot | build #445 of debian wheezy - memcheck is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/debian%20wheezy%20-%20memcheck/builds/445 blamelist: lambday <heavensdevil6909@gmail.com>, Soumyajit De <heavensdevil6909@gmail.com> | 21:39 |
@lisitsyn | iglesiasg: less visible member variables I think | 21:39 |
iglesiasg | lisitsyn: what I am wondering is if it has some impact on swig the attributes | 21:40 |
iglesiasg | the public members definitely do | 21:40 |
@lisitsyn | iglesiasg: I think it handles whatever he sees | 21:41 |
iglesiasg | but the attributes are "always" private | 21:41 |
@lisitsyn | iglesiasg: plus | 21:41 |
@lisitsyn | we can't have any method implementations in headers this way | 21:41 |
@lisitsyn | I mean for getters | 21:41 |
@lisitsyn | you just don't see the class outside of .cpp | 21:41 |
-!- HeikoS [~heiko@12.104.145.5] has quit [Quit: Leaving.] | 21:42 | |
iglesiasg | lisitsyn: so I think the problem with swig is the length of this modshogunPYTHON_wrap.cxx | 21:42 |
@lisitsyn | iglesiasg: yeah sure | 21:42 |
iglesiasg | when we have to compile that file we need a few gb of ram | 21:43 |
iglesiasg | does that file include info about the private members? | 21:43 |
iglesiasg | even if they appear in the header | 21:43 |
iglesiasg | I think it doesn't | 21:43 |
@lisitsyn | iglesiasg: no | 21:43 |
@lisitsyn | iglesiasg: I am curious whether we can use C but not C++ for wrapper as well | 21:44 |
@lisitsyn | probably no | 21:44 |
iglesiasg | lunch time :) | 21:45 |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has joined #shogun | 21:47 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has quit [Client Quit] | 21:48 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has joined #shogun | 21:48 | |
-!- iglesiasg [~iglesias@12.104.145.5] has quit [Ping timeout: 255 seconds] | 21:49 | |
-!- iglesiasg [~iglesias@12.104.145.5] has joined #shogun | 22:03 | |
-!- HeikoS [~heiko@12.104.145.5] has joined #shogun | 22:22 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 22:22 | |
-!- wiking_ [~wiking@info2k1.hu] has joined #shogun | 23:00 | |
-!- wiking [~wiking@huwico/staff/wiking] has quit [Read error: Connection reset by peer] | 23:00 | |
-!- wiking_ [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:00 | |
@HeikoS | lisitsyn: https://github.com/shogun-toolbox/shogun/wiki/SWIG-issues | 23:01 |
@HeikoS | lisitsyn: d-pointers will not solve the swig issues | 23:01 |
iglesiasg | HeikoS: awesome documentation | 23:02 |
@HeikoS | iglesiasg: thanks, feel free to clarify things | 23:02 |
@HeikoS | iglesiasg: currently producing an example | 23:02 |
@HeikoS | iglesiasg: found a relatively nice way | 23:02 |
@HeikoS | iglesiasg: i tried it forlibsvmfil | 23:04 |
@HeikoS | iglesiasg: and it saves 20k lines of code in the swig output | 23:04 |
@HeikoS | !!! | 23:04 |
iglesiasg | fuck, that sounds great | 23:04 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 23:05 | |
shogun-notifier- | shogun: Thoralf Klein :develop * 15209a2 / src/shogun/base/SGObject.cpp: https://github.com/shogun-toolbox/shogun/commit/15209a2c5e584a2044238f11a664c8cf1a8bfcc1 | 23:05 |
shogun-notifier- | shogun: Don't include "class_list" from same directory. | 23:05 |
shogun-notifier- | shogun: Thoralf Klein :develop * 7e15762 / examples/undocumented/libshogun/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/7e15762c72828fbcb89cf7d4c18beaf057fff237 | 23:05 |
shogun-notifier- | shogun: Don't hardcode list of libshogun examples (instead use file glob feature of cmake). | 23:05 |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:05 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:06 | |
shogun-notifier- | shogun: Thoralf Klein :develop * a62e924 / examples/undocumented/libshogun/CMakeLists.txt: https://github.com/shogun-toolbox/shogun/commit/a62e92456bc4f6852d49f85f025d8edb61403b2e | 23:09 |
shogun-notifier- | shogun: Revert "Don't hardcode list of libshogun examples (instead use file glob feature of cmake)." | 23:09 |
shogun-notifier- | shogun: | 23:09 |
shogun-notifier- | shogun: This reverts commit 7e15762c72828fbcb89cf7d4c18beaf057fff237. | 23:09 |
shogun-notifier- | shogun: Thoralf Klein :develop * d13b386 / src/shogun/base/SGObject.cpp: https://github.com/shogun-toolbox/shogun/commit/d13b386b9594d371b4a710fe61b46e06641211e1 | 23:09 |
shogun-notifier- | shogun: Revert "Don't include "class_list" from same directory." | 23:09 |
shogun-notifier- | shogun: | 23:09 |
shogun-notifier- | shogun: This reverts commit 15209a2c5e584a2044238f11a664c8cf1a8bfcc1. | 23:09 |
@HeikoS | iglesiasg: check again | 23:12 |
@HeikoS | https://github.com/shogun-toolbox/shogun/wiki/SWIG-issues | 23:12 |
-!- travis-ci [~travis-ci@ec2-54-80-122-236.compute-1.amazonaws.com] has joined #shogun | 23:15 | |
travis-ci | it's Thoralf Klein'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/39038117 | 23:15 |
-!- travis-ci [~travis-ci@ec2-54-80-122-236.compute-1.amazonaws.com] has left #shogun [] | 23:15 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:16 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:16 | |
iglesiasg | HeikoS: just added to that wiki page the RealFeatures(CSVFile) example to clarify the idea | 23:16 |
@HeikoS | iglesiasg: go for it | 23:17 |
@HeikoS | iglesiasg: yeah why extra interface classes wont help | 23:17 |
iglesiasg | HeikoS: I think I got the idea why it won't work | 23:17 |
iglesiasg | check it if you have a second and delete if I got it wrong ;D | 23:18 |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has quit [Remote host closed the connection] | 23:19 | |
-!- shaochuan [~shaochuan@c-50-184-81-180.hsd1.ca.comcast.net] has joined #shogun | 23:20 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:21 | |
@lisitsyn | HeikoS: ok :) | 23:26 |
@HeikoS | iglesiasg: I am now checking SGVector | 23:26 |
-!- wiking_ [~wiking@info2k1.hu] has joined #shogun | 23:27 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:27 | |
@HeikoS | iglesiasg, lisitsyn. soon pushing these single example for everyone to look/discuss | 23:27 |
@HeikoS | and then if this is good, we should go through thoralfs hit list | 23:27 |
-!- wiking_ [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:27 | |
@lisitsyn | HeikoS: well what d-pointers help is hiding more methods | 23:27 |
@HeikoS | lisitsyn: yeah | 23:27 |
@HeikoS | lisitsyn: and I love it | 23:27 |
@HeikoS | lisitsyn: but not helping for swig, at least we think | 23:28 |
@lisitsyn | HeikoS: yeah sure I don't say it is for swig | 23:28 |
@HeikoS | lisitsyn: but I would love to have this | 23:28 |
@HeikoS | lisitsyn: then we can get more flexible with methods having non shogun parmaeters etc | 23:28 |
@HeikoS | lisitsyn: but for now want to fix travis, which fails most times | 23:29 |
@HeikoS | lisitsyn: could you fix some warnings? | 23:29 |
@HeikoS | lisitsyn: for release? | 23:29 |
@lisitsyn | HeikoS: yes I'll try tomorrow | 23:29 |
@HeikoS | lisitsyn: btw | 23:29 |
@HeikoS | lisitsyn: we talked a lot about refactoring shoguns internals and do proper software engineering | 23:29 |
@HeikoS | lisitsyn: separate algorihtms and core framework | 23:29 |
@HeikoS | lisitsyn: maybe dropping parameter interface | 23:30 |
@HeikoS | lisitsyn: etc | 23:30 |
@HeikoS | lisitsyn: well write it down soon | 23:30 |
@HeikoS | lisitsyn: aer is also cool | 23:30 |
@lisitsyn | HeikoS: ok cool | 23:30 |
@HeikoS | lisitsyn: but thoralf suggested to drop parameters maybe completely, we should definitely talk about this | 23:30 |
@HeikoS | lisitsyn: I am sure we can get shogun pretty light weight using this | 23:30 |
@lisitsyn | HeikoS: if we can drop parameters at all it could be cool | 23:30 |
@lisitsyn | but how? | 23:30 |
@HeikoS | lisitsyn: thoralf also currently builds a script that allow to only compile a certain number of classes | 23:30 |
@HeikoS | lisitsyn: we loose serialisation, model-selection (might drop anyway), clone and equals | 23:31 |
@HeikoS | lisitsyn: alternative: do it properly without all this overhead in the base classes | 23:31 |
@HeikoS | lisitsyn: it really blows up compile time I think, but we would have to investigate that | 23:31 |
@lisitsyn | HeikoS: ehmm we have say svm's C | 23:31 |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:32 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:32 | |
@lisitsyn | where to put that? | 23:32 |
@HeikoS | lisitsyn: keep x-validation | 23:32 |
@HeikoS | then users have to write a loop | 23:32 |
@lisitsyn | no I mean | 23:32 |
@lisitsyn | you have some parameters anyway | 23:32 |
@lisitsyn | ahh you mean SG_ADD thing | 23:32 |
@lisitsyn | not state at all | 23:32 |
@HeikoS | for c in Cs: | 23:32 |
@HeikoS | svm.set_c(c) | 23:32 |
@HeikoS | xval.eval(svm) | 23:32 |
-!- lambday [67157f4f@gateway/web/freenode/ip.103.21.127.79] has quit [Ping timeout: 246 seconds] | 23:33 | |
@HeikoS | lisitsyn: yeah drop that maybe | 23:33 |
@lisitsyn | HeikoS: haha what if we drop state at all :D | 23:33 |
@HeikoS | lisitsyn: rather support fast data IO | 23:33 |
@HeikoS | so that people can store output | 23:33 |
@HeikoS | but dont know yet | 23:33 |
@HeikoS | also drop integration tests | 23:33 |
@HeikoS | and all un-tested classes (or write unit tests) | 23:33 |
@lisitsyn | why? | 23:33 |
@HeikoS | lisitsyn: just overhead | 23:34 |
@HeikoS | lisitsyn: not a lot of use, | 23:34 |
@HeikoS | lisitsyn: but thats for later | 23:34 |
@lisitsyn | well | 23:34 |
@HeikoS | lisitsyn: for now: release and swig | 23:34 |
@lisitsyn | they are helpful sometimes | 23:34 |
@HeikoS | lisitsyn: cant remember a single case where they helped anyone actually | 23:34 |
@HeikoS | lisitsyn: but later | 23:34 |
@lisitsyn | yeah ok | 23:34 |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:37 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:37 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:42 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:42 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:47 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:47 | |
-!- wiking [~wiking@info2k1.hu] has joined #shogun | 23:57 | |
-!- wiking [~wiking@info2k1.hu] has quit [Read error: Connection reset by peer] | 23:57 | |
--- Log closed Sun Oct 26 00:00:40 2014 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!