Open in new window / Try shogun cloud
--- Log opened Thu Sep 27 00:00:17 2012
-!- in3xes_ [~in3xes@] has quit [Ping timeout: 252 seconds]01:36
shogun-buildbotbuild #115 of nightly_default is complete: Failure [failed test]  Build details are at
-!- blackburn [~blackburn@] has joined #shogun07:43
-!- hoijui [] has joined #shogun08:06
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]08:33
-!- hoijui [] has quit [Ping timeout: 265 seconds]12:40
-!- heiko1 [] has joined #shogun13:55
-!- heiko1 [] has quit [Ping timeout: 245 seconds]14:29
-!- blackburn [~blackburn@] has joined #shogun14:46
-!- heiko1 [] has joined #shogun16:00
-!- heiko1 [] has quit [Ping timeout: 248 seconds]16:24
-!- heiko1 [] has joined #shogun16:37
heiko1blackburn around?16:57
blackburnheiko1: hey, yes16:57
heiko1I have another problem with c++ (many these days)16:57
heiko1I want to make a subclass of a templated class16:57
heiko1in particular, subclass of CSTreamingDeanseFeatures<T>16:58
heiko1template <class T>16:58
heiko1class CStreamingDenseDataGenerator: public CStreamingDenseFeatures<T>16:58
heiko1but that doesnt work16:58
heiko1a) I cannot access members of CStreamingDenseFeatures in methods of CStreamingDenseDataGenerator<T>16:58
heiko1b) shogun doesnt recognise the new class as generic so compiling class_list.cpp fails16:58
heiko1any ideas?16:58
heiko1blackburn, weird problem16:59
blackburnyeah pretty weird17:01
heiko1man no idea  :)(17:03
-!- ptizoom [55d25a64@gateway/web/freenode/ip.] has joined #shogun17:03
blackburnheiko1: about a) - does it compile?17:04
heiko1only b) fails to compile17:05
heiko1oh sorry17:05
heiko1complains that the member elements are not known17:05
heiko1ah but I can avoid this using17:05
heiko1still b)17:05
heiko1in class list, generic classes are instanciated via this switch thing17:06
heiko1that is automatically generated for generic classes17:06
heiko1but for this subclass, htis doesnt happen17:06
blackburndid you instantiate your new class?17:07
blackburnin cpp17:07
ptizoomhi, I am looking into associating PDL and Shogun (by replacing numpy). I am progressing slowly...I have some framework (which is not yet compiling), but is it possible to have a development git-branch ?17:09
blackburnptizoom: I am unsure I got what you mean17:10
blackburndev git branch of shogun?17:11
ptizoomblackburn: is shogun under the git?17:11
blackburnwell, yes17:12
blackburnbut you may commit everything you want in your fork17:12
heiko1blackburn, how do you mean that?17:13
ptizoomis it easy to get a writting branch from the project?17:13
heiko1ptizoom, that takes a while, but as blackburn said, you can just push everything into your fork and then send pull-requests. We then review these, give feedback, and evtl. merge17:14
heiko1blackburn, its an error in the script17:17
heiko1I tuned it manually and now it compiles17:17
blackburnptizoom: yeah pretty easy17:20
blackburnah heiko1 answered17:20
blackburnheiko1: how does check whether it is a template?17:21
heiko1blackburn, currently checking17:21
ptizoomheiko1: cool17:22
heiko1blackburn, is there any problem with subclassing templated classes?17:37
heiko1I not get weird errors17:38
heiko1for example when trying to create an instance I get17:38
heiko1undefined reference to `shogun::CStreamingDenseDataGenerator<double>::CStreamingDenseDataGenerator()'17:38
heiko1which is totally weird since the definition is there17:38
heiko1why is c++ alway such a horror when it comes to certain things17:39
heiko1this is so annoying, I am wasting my time with this stupid stuff where all I want is a generic subclass17:40
heiko1cant be that difficult17:40
blackburnso it fails to find implementation?17:40
blackburncan you paste the code somewhere?17:41
heiko1yes, will do17:41
heiko1first error you get should be class_list.cpp17:43
heiko1I think its related to pure virtual functions17:45
heiko1and that inheritance of templated classes doesn not give superclasses elements17:46
heiko1(which would explain the pure virtual methods and the non-accessability of superclass members)17:46
heiko1blackburn, any ideas?17:47
blackburna minute17:47
heiko1blackburn, leaving now, if you would find something that was amazing :)17:59
blackburnheiko1: sorry got distracted17:59
heiko1no worries17:59
blackburnlet me check now, I'll let you know17:59
heiko1cool, thx18:00
heiko1ill check back in 15 mins in case something is missing to test it18:00
blackburnheiko1: I think you have to instantiate using18:01
blackburntemplate class CStreamingDenseDataGenerator<bool>;18:01
blackburnin .cpp18:02
blackburnsee DenseFeatures.cpp for example18:02
-!- audy [] has quit [Changing host]18:11
-!- audy [~audy@unaffiliated/audy] has joined #shogun18:11
heiko1blackburn, no that doesnt work18:16
heiko1have you tried it?18:17
heiko1blackburn, compiler complains about non-pure virtual methods18:17
blackburnhmm do you have some pure virtual methodds around?18:18
blackburnI can't see any18:19
-!- zxtx [] has joined #shogun18:20
heiko1it complains about all the pure virtual methods from CStreamingFetaures that are imeplemented in CStreamingDenseFeatures18:21
heiko1but these are for some reason not visible in the subclass18:21
blackburnheiko1: did you try forward declaration of streaming dense feats?18:23
heiko1blackburn, what is that?18:23
blackburntemplate <class ST> class CStreamingDenseFeatures;18:24
blackburnbefore definition in *.h18:24
heiko1ah sorry ok, Ill try18:24
blackburnyeah C++ is pain..18:25
heiko1blackburn, no doesnt change anything18:25
heiko1but makes sense since the header is included above18:25
heiko1yeah man, c++ and OO is so annoying sometimes18:25
heiko1no interfaces, no generics18:25
heiko1I need these all the time :D18:25
heiko1blackburn, Ill update the listings, have you tried to compile them?18:26
blackburnno, not yet18:26
blackburnlet me try then18:26
heiko1would be nice ofyou18:26
blackburnbase/class_list.cpp:414:103: error: expected type-specifier before 'CStreamingDenseDataGenerator'18:29
blackburnis that you get?18:29
heiko1you can get around that18:30
heiko1via searching for CStreamingDenseDataGenerator in class_list.cpp18:30
heiko1and then cut the line and replace it by a block that is similar to the other templated classes, e.g. CDenseFeatures18:31
heiko1with a switch over the types18:31
blackburnheh yeah but we should rather fix class_list thing18:31
blackburn watch that video meanwhile18:32
heiko1it doesnt solve the problem untimately18:32
heiko1cut the line18:32
heiko1static CSGObject* __new_CStreamingDenseDataGenerator(EPrimitiveType g) { return g == PT_NOT_GENERIC? new CStreamingDenseDataGenerator(): NULL; }18:32
heiko1and replace it by a block that you can find from line 753 ...18:32
heiko1lol the video is great .)18:33
blackburnbut why does it treat your class as not templated?18:33
heiko1I think its because of the templates doing something wrong18:34
heiko1blackburn, there is another problem even before the class_list18:35
heiko1if you add18:35
heiko1template class CStreamingDenseDataGenerator<bool>; in the last line of the .cpp file18:36
heiko1it doesnt compile already18:36
blackburnlol! you've got to make it one line18:36
heiko1due to these pure virtual function18:36
blackburn 32 template <class T> class CStreamingDenseDataGenerator: public CStreamingDenseFeatures<T>18:36
blackburnthis way18:36
blackburnnot two liner18:36
blackburnclass_list solved then18:37
blackburnokay lets get to undefined reference problem18:37
heiko1since python is line based18:37
blackburnnot a python but the thing18:37
heiko1do you now also get the undefined references?18:37
heiko1These can be solved by adding the lines18:38
blackburnwork work sleep then work18:38
heiko1template class CStreamingDenseDataGenerator<bool>;18:38
heiko1but then another error appears18:38
blackburnokay let me get to next error in our funny journey18:38
heiko1what do you think of this one?18:39
blackburnfeatures/StreamingDenseDataGenerator.cpp:16:35: error: type 'shogun::CStreamingDenseDataGenerator<bool>' is not a direct base of 'shogun::CStreamingDenseDataGenerator<bool>'18:39
blackburnokay do not blame C++, it is you :)18:40
blackburncheck your constructor carefully18:40
blackburnyou kinda calling it itself again18:40
blackburnhowever other errors stay :D18:41
heiko1i already fixed the others18:41
heiko1copy paste -.-18:41
heiko1thanks a lot18:41
heiko1guess i have been starring to the monitor too much18:41
blackburnlets dance north korean style18:41
blackburnkim jong style18:41
heiko1would never have found this18:41
heiko1yeah, go :D18:41
heiko1yep works now18:42
blackburnI absolutely love this country btw, would love to visit they18:42
blackburnbut too scary18:42
heiko1why do you love it?18:42
blackburnit is like they freezed soviet union in 60s18:43
blackburnso funny and crazy18:43
heiko1soyou like that retroish style :)18:43
heiko1you should get a lomography camera then :)18:44
blackburnhah one friend of mine has a few18:44
heiko1my girlfriend has a a few18:44
heiko1they make cool phoos18:44
blackburnhis photos look like this:
heiko1there are more errors18:45
heiko1inline-Funktion >>shogun::EFeatureType shogun::CStreamingDenseFeatures<T>::get_feature_type() const [with T = long double]<< verwendet, aber nirgendwo definiert18:45
heiko1bla used but not defined18:45
blackburnthat's why you didn't paste18:46
heiko1and these are exactly the pure virtual methods that are imeplemented in CStreamingDenseFeatures18:46
heiko1you got it :)18:46
blackburnaber nirgendwo definiert!18:46
heiko1thats: but nowhere defined :)18:47
blackburnI know only what GmbH is18:47
blackburngesselschaft mit beschraenkter haftung!18:47
heiko1what is also strange: have a look at CStreamingDenseFeatures18:48
blackburnobschestvo s ogranichennoy otvetstvennostyu18:48
heiko1there is an element current_sgvector18:48
blackburnnot that shorter in russian, lol18:48
heiko1but I cannot access that from my new class, compile error18:48
heiko1Russion looks mess18:48
heiko1and its beschr?????nkter )18:48
blackburnyeah, I thought ae works too18:48
heiko1yeah it does18:49
blackburnlets see updated code then18:49
heiko1I will create another pastebin18:50
blackburnheiko1: btw I am now developing the ideal machine learning environment (TM)18:50
blackburncheck one idea18:50
heiko1Ill check18:51
heiko1btw I renamed the variable to current_vector in my code-base18:51
blackburnbriefly - all the things that happen to 'context' variables are going to be serialized18:51
blackburnbut that's all transparent and you don't care about that18:52
heiko1but thats python only or?18:52
blackburnyeah sure18:53
blackburnsuch madness can't happen in any normal language18:53
blackburndo you like the idea?18:53
blackburnmy other extension is18:54
blackburnwith operation('reading files') as context:context.train_data = numpy.loadtxt('train.csv',delimiter=",",skiprows=1)context.test_data = numpy.loadtxt('test.csv',delimiter=",",skiprows=1)18:54
blackburnwith operation('reading files') as context:18:54
blackburncontext.train_data = numpy.loadtxt('train.csv',delimiter=",",skiprows=1)18:54
blackburncontext.test_data = numpy.loadtxt('test.csv',delimiter=",",skiprows=1)18:54
blackburncontext.train_data is going to be saved and then serialized to file18:54
blackburnafter with block18:54
blackburnI am thinking about a way to load it from serialized automatically as well18:55
heiko1yeah I mean, thats all the magic of pickle18:55
heiko1nice, like it18:55
blackburnyes, it is easy18:55
blackburnthe thing I am trying to reach18:55
heiko1though its not really shogun-ish since only works in one binding18:55
blackburnis transparency of such operations like measuring time18:55
blackburnor serialization18:56
blackburnor rather implicit way18:56
blackburnI mean I want to save everything without calling anything18:56
heiko1yeah, thats great use18:56
blackburnhave you seen patterns for ML article?18:57
heiko1where ?18:57
heiko1doing that all the time anyway :)18:58
blackburnyes, thats the spirit18:58
blackburnfeatures/streaming/StreamingDenseDataGenerator.cpp:75:68: error: invalid conversion from 'const char*' to 'int32_t {aka int}' [-fpermissive]18:58
blackburn../shogun/base/Parameter.h:206:21: error:   initializing argument 1 of 'shogun::TParameter* shogun::Parameter::get_parameter(int32_t)' [-fpermis18:58
blackburnthat's what I get18:59
heiko1ah year18:59
heiko1I added stuff19:00
heiko1wait, I will push that quicklly19:00
blackburnheiko1: btw are you using vim?19:00
heiko1no, why?19:00
blackburnjust curious19:00
blackburnI argue with my team lead at job every day :D19:00
heiko1blackburn, you can just comment that line  7519:01
blackburnI am a vim guy, he says sth like GUI rocks19:01
heiko1no I actually use eclipse19:01
heiko1like browsing around in the code by mouse19:01
blackburnI see19:01
blackburnlooks like it appears to compile then19:02
heiko1but change something and compile again:19:02
blackburnyeah, I get quite a lot warnings19:03
heiko1these are strange19:03
heiko1because the methods are all implemented in superclass19:03
heiko1(in fact, thats why I derived)19:03
blackburnprobably we have some other glitch around there19:04
blackburngangnam style!19:05
blackburnwoop woop woop woop19:05
heiko1code works apart from these warnings btw :)19:07
blackburncan't see nothing wrong19:07
blackburnwalking kim jong style19:07
-!- n4nd0 [] has joined #shogun19:22
n4nd0wiking: around?19:31
blackburnn4nd0 entered room kim jong style19:31
n4nd0haha why?19:32
n4nd0blackburn: any news about your interview?19:40
blackburnn4nd0: no, nothing yet19:40
blackburnthey have to come up with a reject message :D19:41
n4nd0I bet you don't believe that ;)19:41
blackburnno I really believe I am not that cool :)19:43
-!- heiko1 [] has left #shogun []19:49
wikingn4nd0: yes21:14
n4nd0wiking: how is it going using the SO framework?21:15
n4nd0blackburn: do you know if the post-gsoc post came eventually to life?21:20
n4nd0post-gsoc post haha21:20
n4nd0funny name21:20
blackburnyou mean t-shirt thing?21:20
blackburniirc it should arrive before 30th21:20
n4nd0no no21:21
n4nd0the post21:22
n4nd0sonney2k asked for help for a post and I sent him by mail one21:22
n4nd0a couple of weeks ago he said he'd started fixing it21:22
wikingn4nd0: i've got convergence with bmrm21:22
blackburnwell no idea21:24
n4nd0wiking: really? I think you should actually get the same solution as with PrimalMosek21:28
wikingmy problem is/was really the memory consumption22:09
wikingof mosek22:09
n4nd0oh sorry I misunderstood. for some strange reason I thought I read that it didn't converge with bmrm22:18
n4nd0I don't know if I screw sth up with my code using mosek22:18
-!- zxtx [] has quit [Ping timeout: 240 seconds]22:26
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]22:43
-!- n4nd0 [] has quit [Ping timeout: 248 seconds]23:02
--- Log closed Fri Sep 28 00:00:17 2012