Open in new window / Try shogun cloud
--- Log opened Sat Jul 09 00:00:14 2016
-!- lambday [56a397c5@gateway/web/freenode/ip.] has quit [Ping timeout: 250 seconds]01:39
shogun-buildbotbuild #28 of clang - thread analysis is complete: Failure [failed compile]  Build details are at  blamelist: Heiko Strathmann <>, Saurabh7 <>03:44
shogun-buildbotbuild #26 of clang - undefined behaviour analysis is complete: Failure [failed compile]  Build details are at  blamelist: Heiko Strathmann <>, Saurabh7 <>03:48
shogun-buildbotbuild #42 of memleak - valgrind is complete: Failure [failed memory check]  Build details are at  blamelist: Heiko Strathmann <>, Saurabh7 <>05:38
shogun-buildbotbuild #1040 of nightly_none is complete: Failure [failed compile]  Build details are at  blamelist: Heiko Strathmann <>, Saurabh7 <>05:48
shogun-buildbotbuild #1170 of nightly_default is complete: Failure [failed test notebooks]  Build details are at  blamelist: Heiko Strathmann <>, Saurabh7 <>06:35
-!- mizari [] has joined #shogun07:34
-!- mizari [] has quit [Quit: Leaving]11:55
-!- sanuj [~sanuj@] has joined #shogun12:40
sanujbzdmgz, there?12:40
-!- sanuj [~sanuj@] has quit [Ping timeout: 264 seconds]13:26
-!- sanuj [~sanuj@] has joined #shogun13:39
-!- sanuj [~sanuj@] has quit [Quit: Leaving]14:13
-!- sanuj [~sanuj@] has joined #shogun14:13
sanujbzdmgz, there?14:13
bzdmgzsanuj: FINALLY yes14:15
bzdmgzjust fixed some crazy thing14:15
sanujbzdmgz, office work on weekend?14:15
bzdmgzno but fixing it14:15
bzdmgzcrazily out of SLA14:15
bzdmgzok seen your mail14:16
bzdmgzlet me check14:16
bzdmgzI agree with heiko but we need to separare those14:17
bzdmgzI mean we need that list14:17
bzdmgzbut we can't combine list and create14:17
bzdmgzwhat about removing swig part YET14:17
bzdmgzwe tried, it works14:17
bzdmgzlet me put that so heiko knows14:18
sanujbzdmgz, i'll remove swig from create() and update the PR14:18
bzdmgzyeah tanks14:18
sanujbzdmgz, see the other PR also14:19
bzdmgzsanuj: yeah14:19
bzdmgzsanuj: actually creating14:19
bzdmgzlike clone14:19
bzdmgzand equals14:19
bzdmgzmaybe it makes sense?14:19
sanujbzdmgz, are you talking about jinja?14:20
bzdmgzsanuj: I tend to think it makes sense14:20
sanuji've done like the jinja generated unit-tests only14:20
bzdmgzwhat do you think?14:20
sanujjinja is good14:20
bzdmgzno I mean14:20
bzdmgzjust copy what happens in14:20
bzdmgzclone unit test14:20
bzdmgzto test all the clasess are created14:21
sanujI see14:21
sanujbzdmgz, but we need to decide what classes to add14:21
bzdmgzthe only thing bothers me14:21
bzdmgzis that you use14:21
bzdmgzok they all have base class14:21
bzdmgzjust use create<CSGObject>14:22
sanujbzdmgz, see my updated MockObject14:22
sanuji have created a pointer of CKernel* there14:22
bzdmgzwhere where14:22
sanujin the jinja pr14:22
bzdmgzok let me check14:23
bzdmgzsanuj: oh that looks cool14:26
sanujbzdmgz, the PR?14:26
bzdmgzjinja thing look fancy14:26
sanujbzdmgz, R is failing14:27
sanujdid you see my last comment14:27
bzdmgznot sure14:27
bzdmgzI think it just doesn't support overloading?14:27
bzdmgzsanuj: we need a patch to disable it for R then14:27
sanujbzdmgz, does anyone use shogun in R?14:28
bzdmgzdoes anyone use shogun14:28
bzdmgz:D :D JK14:28
bzdmgzsanuj: no idea14:28
bzdmgzwe do not track that14:28
bzdmgzshould intergrate some spyware14:28
bzdmgzsanuj: just ignore it do not let it stop you :D14:29
bzdmgzI mean some #ifdef maybe14:29
sanuji see14:29
bzdmgzyou can find some examples actually14:29
bzdmgzin our .i files14:29
sanujthe error happens because of this line in modshogun.R14:29
sanujvaccessors = c('io', 'parallel', 'version', 'm_parameters', 'm_model_selection_parameters', 'm_gradient_parameters', 'm_hash', 'get''get''get''get''get''get''get''get');14:29
sanujthe 'get' madness14:29
sanujbzdmgz, some similar issues were reported by14:30
bzdmgzyeah I've seen14:31
bzdmgzshit happens so we have to avoid that14:31
sanujbzdmgz, so no tags-swig for R?14:31
sanujwill use ifdef as you suggested14:31
bzdmgzunless it is fixed in swig14:31
sanujbzdmgz, i think will take time14:31
sanujif they fix it14:32
bzdmgzlets just wait14:32
bzdmgzonce fixed we enable it for R14:32
sanujbzdmgz, look at this
sanujif i use CGaussianKernel* instead14:33
sanujit gives error from because the type are not same as i use setKernel() which expects a CKernel*14:34
bzdmgzshouldnt be casted?14:35
sanujbzdmgz, will try casting14:35
bzdmgzno I mean implicitly14:35
sanujbzdmgz, well if it was implicitly casted, then it would have worked14:36
sanujbzdmgz, if we do there a problem?14:37
bzdmgzI am not sure what's the problem14:37
sanujcool :D14:38
sanuji was just confirming14:38
sanujbzdmgz, i can't add classes till we decide what classes to add14:38
sanujbzdmgz, so what are the next steps for tags14:39
sanujand for plugins14:39
sanuji'll update the create PR14:39
bzdmgzsanuj: ok it is hard to decide what classes14:40
sanujand the jinja PR to handle R error14:40
bzdmgzsanuj: what about selecting just a few14:40
sanujbzdmgz, yes14:40
bzdmgzjust to make the simplest complete example work14:40
bzdmgzclassification with kernel svm14:40
sanujbzdmgz, okay14:40
bzdmgzwhole thing should be created by create14:40
bzdmgzno `new` involved14:41
sanujbzdmgz, i will have to change internals of classes to support tag-parameters14:41
sanujbzdmgz, are you talking about using tag-parameters for kernel svm classification example?14:42
sanujor just create?14:42
bzdmgzsanuj: I think tag is tougher14:45
bzdmgzjust create firt14:45
sanujbzdmgz, so i think the bottleneck is to decide what classes to add for tags14:45
sanujbzdmgz, and it is better if you and heiko do it14:46
bzdmgzit is nearly impossible14:46
bzdmgzwe can start with something and iterate14:46
sanujbzdmgz, okay so what is the next step for the tags PR after i fix R issues14:47
bzdmgzsanuj: I have to think about it14:47
bzdmgzthere are quite a few things to do14:47
bzdmgznot sure how to decompose14:47
bzdmgzbut it is moving quite good14:48
bzdmgzwe will have something fancy quite soon14:48
sanujbzdmgz, next thing for plugins, after i have the svm example ready with create()14:48
bzdmgzsanuj: dlopen thing is due I guess14:49
sanujjust asking this now.....because i don't know when you will be available on irc next :D14:49
sanujbzdmgz, so migrate aer code in shogun14:49
bzdmgzI will try to follow up next week14:49
bzdmgzthis week was crazy14:49
sanujhaha i see14:49
bzdmgzsanuj: so the basic idea14:50
bzdmgzas you have create by string14:50
bzdmgzmake it able to call some library14:50
bzdmgzto ask it to create this object14:50
sanujbzdmgz, so it is like using create with your aer plugin code14:51
sanuji'll have to modify create again then14:51
bzdmgzbut inside of create14:51
bzdmgzno interface involved14:51
sanujbzdmgz, what interface?14:52
bzdmgzof this 'create' function14:52
sanujabstract class14:52
bzdmgzI mean from user experience it is still create14:52
bzdmgzthe only difference is how we load it14:52
sanuji see14:52
sanujso from user point of view, create() doesn't change14:52
sanujonly it's internal implementation changes14:53
sanujbzdmgz, so i'll first make this create svm example work14:53
sanujbzdmgz, and migrate your aer plugin code on a different branch14:53
bzdmgzyes makes sense14:53
sanujand make it work with create()14:53
sanujbzdmgz, but for this i think i would also need to write a plugin14:54
sanujdummy plugin14:54
bzdmgzgaussian kerel14:54
sanujbzdmgz, so what's the point of merging create in develop right now?14:55
bzdmgzdid I say that?14:56
sanujokay, not merging then?14:56
bzdmgzI think it can go to tags14:56
sanujbzdmgz, will send a pr to tags14:57
-!- sanuj [~sanuj@] has quit [Ping timeout: 264 seconds]15:07
-!- Saurabh7 [Saurabh7@gateway/shell/panicbnc/x-xohcuxorptqneppp] has quit [Ping timeout: 264 seconds]15:10
-!- Saurabh7 [Saurabh7@gateway/shell/panicbnc/x-qqkcvttzzqzpgvhm] has joined #shogun15:31
-!- HeikoS [] has joined #shogun15:41
-!- mode/#shogun [+o HeikoS] by ChanServ15:41
-!- HeikoS [] has quit [Quit: Leaving.]15:53
-!- sanuj [~sanuj@] has joined #shogun16:15
-!- sanuj [~sanuj@] has quit [Ping timeout: 272 seconds]16:53
-!- sanuj [~sanuj@] has joined #shogun17:43
-!- HeikoS [] has joined #shogun19:25
-!- mode/#shogun [+o HeikoS] by ChanServ19:25
-!- HeikoS [] has quit [Quit: Leaving.]19:34
sanujwiking, there?19:59
sanujbzdmgz, there?20:18
bzdmgzsanuj: yeap20:18
sanujbzdmgz, i have a cmake question20:18
sanujJINJA2_FOUND is not set20:18
sanujby FIND(Jinja2)20:19
sanujinstead there is this variable20:19
sanuji was trying to do this20:20
sanuj# Generate automatic shogun-base.i from jinja2 templates20:20
sanujADD_CUSTOM_COMMAND(OUTPUT shogun-base.i20:20
sanuj    COMMAND ${PYTHON_EXECUTABLE} ${COMMON_MODULAR_SRC_DIR}shogun-base.i.py20:20
sanuj    ${COMMON_MODULAR_SRC_DIR}shogun-base.i.jinja220:20
sanuj    shogun-base.i20:20
sanuj    ${COMMON_MODULAR_SRC_DIR}shogun-base-list.txt20:20
sanuj    DEPENDS ${COMMON_MODULAR_SRC_DIR}shogun-base.i.py20:20
sanuj    ${COMMON_MODULAR_SRC_DIR}shogun-base.i.jinja220:20
sanuj    COMMENT "Generating shogun-base.i")20:20
sanujbzdmgz, ^20:20
bzdmgzyeah it is not being set20:21
bzdmgzyou can find it in cmake/FindJinja2.cmake20:21
bzdmgzsanuj: I think you can patch it20:21
bzdmgzfind_package_handle_standard_args(LibXslt FOUND_VAR LibXslt_FOUND20:21
bzdmgz                                         REQUIRED_VARS LibXslt_LIBRARIES LibXslt_INCLUDE_DIRS20:21
bzdmgz                                         VERSION_VAR LibXslt_VERSION_STRING)20:21
bzdmgzyou can add FOUND_VAR20:21
bzdmgzwith JINJA2_FOUND20:22
bzdmgzjust like in this example20:22
bzdmgzit should work20:22
sanujbzdmgz, it is for this
sanujbzdmgz, thanks20:23
sanujgoingg to sleep20:23
sanujwill do this tomorrow20:23
-!- sanuj [~sanuj@] has quit [Quit: Leaving]20:24
-!- HeikoS [~heiko@] has joined #shogun23:11
-!- mode/#shogun [+o HeikoS] by ChanServ23:11
-!- HeikoS [~heiko@] has quit [Quit: Leaving.]23:52
--- Log closed Sun Jul 10 00:00:16 2016