Open in new window / Try shogun cloud
--- Log opened Mon Aug 29 00:00:20 2011
--- Day changed Mon Aug 29 2011
blackburn55k$ per year?! the default salary in calculator rocks00:00
blackburnsonney2k: lets sell shogun :D00:00
blackburncurrently my salary is 3k/year ahaha00:02
blackburnsonney2k: make in examples currently requires doc for .class files00:05
blackburnin java00:06
blackburndocumentation for "MatrixTest.class" (file "descriptions/modular/MatrixTest.txt") missing00:08
blackburndocumentation for "" (file "descriptions/modular/MatrixTest.txt") missing00:08
@sonney2kerase the class files before doing make doc or fix the to exclude .exe  / .class files00:12
@sonney2kmatrix/vector test should be removed btw00:15
CIA-3shogun: Sergey Lisitsyn master * r2b3ca25 / (2 files): Remove Matrix/VectorTest from java modular examples -
CIA-3shogun: Sergey Lisitsyn master * r28ce82a / (6 files): More doc for examples -
CIA-3shogun: Sergey Lisitsyn master * rfb4ddae / (20 files): Kernel examples doc -
-!- in3xes [~in3xes@] has joined #shogun01:16
shogun-buildbotbuild #182 of libshogun is complete: Failure [failed compile]  Build details are at  blamelist: blackburn91@gmail.com01:34
serialhexHA!!! blackburn GOT BLAMED!!!!!  IT'S UR FAULT!!!!!!01:50
serialhexbtw, is sonney2k around still??01:50
serialhexhmm, asleep probably, i'll e-mail ya the problem i'm getting (same problem, 2 files)01:51
blackburnserialhex: definitely not my fault ;)01:53
serialhexno?? shogun-buildbot says different: "blamelist:"01:53
blackburnyeah but take a look on log01:54
blackburncompiler got segfault hah01:54
serialhexooh, lovely!01:55
serialhexyou just got blamed b/c you were the last to commit, and with a bunch of textfiles01:55
serialhexi would think it would blame everyone who commited between the time it last built & the time it compiled01:57
serialhexhey, do you have python modular compiled & installed??01:59
serialhex^ blackburn01:59
blackburnserialhex: I could compile now01:59
serialhexwell i want to test something, if it's not too much of a pain for you02:00
serialhexthough the ruby would be better02:00
shogun-buildbotbuild #163 of octave_modular is complete: Failure [failed compile]  Build details are at  blamelist: blackburn91@gmail.com02:00
serialhexcause then i could give you the code thats bein a pain for me02:00
blackburnso what to install?02:01
shogun-buildbotbuild #147 of ruby_modular is complete: Failure [failed compile]  Build details are at  blamelist: blackburn91@gmail.com02:01
blackburnbuildbot goes crazy02:02
blackburnserialhex: got python installed, what to do?02:03
serialhexsee if in examples/undoc.../pyth_mod works02:04
serialhex(yes, ui'm lazy)02:04
serialhexwhen i run the ruby version iget a wonky error, and i'm not sure if it's just ruby or if it's shogun in general02:06
blackburnserialhex: yeah it works02:08
blackburnwhat is the error you get?02:08
serialhex...i'll gist it02:09
serialhexmaybe it's soemthing silly i'm missing in the translation between python & ruby....02:10
serialhex^ blackburn02:10
shogun-buildbotbuild #174 of cmdline_static is complete: Exception [exception git]  Build details are at  blamelist: blackburn91@gmail.com02:11
shogun-buildbotbuild #152 of python_modular is complete: Exception [exception git]  Build details are at  blamelist: sonne@debian.org02:11
blackburnserialhex: yeah I know what the error is02:13
serialhex??? please tell!!!02:13
blackburnserialhex: check if your load_dna removes '\n' char02:14
blackburnif yes so it could be typemap error02:14
serialhexhmm, ok02:15
blackburnserialhex: there are 120 lines in fm_test_dna.dat02:17
blackburn92 in train_dna02:17
serialhexhmm, it was joining them nto an array, each line a new element in the array02:17
blackburnso hist[0]=9202:17
blackburnis number of '\n's02:18
blackburndoes it remove \n char?02:18
serialhexyes, it did, but then i changed it to put the entire file into the array, or just use the entire file, nothing :(02:18
blackburnserialhex: try to cut lines02:20
serialhexhmm... something wonky is going on...02:20
blackburne.g. take half of all of them02:20
serialhexoh, derr....02:20
blackburnif it is still failing - I suspect only typemap02:20
blackburnserialhex: so how?02:23
serialhexi forgot, the version of modshogun.rb i'm using is *not* being loaded, so nothing i do in it will fix anything, but i'm going to copy it over and see if i can fix the problem...02:24
blackburnserialhex: but it is ruby load02:25
blackburnnot modshogun.rb related02:25
serialhexwell the modshogun.rb has the load_whatever functions in it02:25
blackburnI see02:26
blackburnserialhex: just make a try with some not loaded data but ACTG02:26
serialhexhmm, ok02:26
blackburnyou can be sure in this way there are no '\n'02:26
serialhexhmm... with or w/o the \n i get a new error with this: traindna = "ACTG"02:28
serialhexdistribution_histogram_modular.rb:16:in `set_features': Wrong arguments for overloaded method 'StringCharFeatures.set_features'.02:29
serialhex^ is trhe new error :P02:29
blackburnserialhex: try array of "ACTG" "TCGA"02:29
blackburn[] or so don't know how it is in ruby02:29
serialhexnope, same error as before with traindna = ["ACTG\n", "TGCA\n"]02:30
blackburnwrong number??02:30
serialhexalso w/o the newlines it dosnt work...02:32
serialhexso i'm thinking the typemap...02:32
blackburnserialhex: so [] things doesn't work anyhow?02:32
serialhexnope, apparently not02:33
blackburnserialhex: so what returns load_dna ?02:34
blackburn[] too02:35
blackburnserialhex: okay sorry but it is 4-35 there already :) I have to sleep02:35
serialhex... and so on02:35
blackburnhmm wait02:35
serialhexok, np... i'll e-mail sonney2k & sploving02:36
blackburnso how can ["ACTG", "TACG"] doesn't02:36
blackburnbut ^ do?02:36
serialhexit dosnt02:36
serialhexnone of them work02:36
blackburnserialhex: I mean when you use load_dna it have different error about alphabet?02:36
blackburnbut if shorter ACTG version it says about wrong num of argz02:37
serialhexnope, when it's not in an array i get a different error, but when it's in an array i get the: [ERROR] ALPHABET too small to contain all symbols in histogram02:37
blackburncan't understand the difference between load_dna result and []02:37
serialhexwith or without newlines02:37
serialhexload_dna result is a longer array02:38
blackburnserialhex: so will ["ACTG", "TACG"] produce alphabet error?02:38
serialhexand so will ["ACTG\n", "TGCA\n"]02:38
blackburnserialhex: how many histogram elements there are with "ACTG" ?02:39
serialhex2 it says02:39
blackburnhist[] =02:40
blackburnhow many records of this form ^02:40
blackburnand with ACTG\n?02:40
blackburnserialhex: I have a fix but you probably have to recompile02:41
serialhexok... i should prob do that anyway with the latest :P02:42
blackburnserialhex: can you please gist script you used for ^?02:42
serialhexthats the latest v of the file & error02:44
blackburnserialhex: the problem is for some *STRANGE* reason baozeng copies len+1 of string chars02:46
blackburnin "in" typemap02:46
serialhexhmm, ok02:46
blackburnserialhex: so ACTG is something 656771840002:47
serialhexok, sounds like a fun error :P02:47
shogun-buildbotbuild #165 of octave_static is complete: Failure [failed compile]  Build details are at  blamelist: blackburn91@gmail.com02:56
blackburnserialhex: got wrong number of args..03:03
serialhexhmm... really?03:03
serialhexyes, it is :-/03:03
blackburnserialhex: w/o \n produces wrong number of args03:03
blackburnw/ - alphabet error03:04
serialhexwell, get to bed, i'll e-mail sonney2k & sploving about it and maybe we can come up with something?03:04
serialhexi dunno :(03:04
blackburnserialhex: let they know that string takes *one* char more than needed03:06
blackburnit is pretty certain03:06
serialhexok, will do03:06
blackburnbut no fix yet03:06
blackburnokay 5-06 time to sleep ahah03:07
blackburnsee you03:07
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]03:08
shogun-buildbotbuild #151 of r_modular is complete: Failure [failed compile]  Build details are at  blamelist: sonne@debian.org03:13
-!- in3xes [~in3xes@] has quit [Remote host closed the connection]03:23
shogun-buildbotbuild #154 of csharp_modular is complete: Failure [failed compile]  Build details are at  blamelist: sonne@debian.org03:40
shogun-buildbotbuild #166 of octave_static is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #164 of octave_modular is complete: Success [build successful]  Build details are at
@sonney2kserialhex, seems like the NULL is in the string05:59
serialhexahh, ok05:59
@sonney2kguess the fix is trivial05:59
@sonney2kjust not use len+1 in the typemap but len06:00
serialhexok cool06:00
CIA-3shogun: Soeren Sonnenburg master * r50aac3e / src/interfaces/ruby_modular/swig_typemaps.i : don't use len+1 for the string but len -
@sonney2kthis ^ should help06:02
serialhexok, i'll pull & try06:03
shogun-buildbotbuild #183 of libshogun is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #163 of python_static is complete: Success [build successful]  Build details are at
@sonney2kserialhex, did you make some progress on examples?06:17
serialhexi've got a bunch... i think like10-20 or something?  a lot of them just won't work though06:18 the preprocesor examples, i can't find the modules :-/06:18
@sonney2kserialhex, what happens?06:20
serialhexwell, like preprocessor_hessianlocallylinearembedding_modular.rb, i can't find a Modshogun::Hessian* anything....06:21
@sonney2kserialhex, then you don't have atlas/blas/lapack installed06:21
serialhexahh, ok06:22
serialhexwhere is that in debian?  just apt-get atlas??06:22
@sonney2kdo apt-get install libblas-dev liblapack-dev libatlas-base-dev06:22
@sonney2kthen recompile and then it will work06:22
serialhexok, recompiling...  my machine is slow so i'll know later in the am06:25
@sonney2kserialhex, btw we had the same issue w/ '\0' being in the string in other $LANG06:27
@sonney2kso I am pretty sure that this fixes06:27
serialhexahh, ok...06:27
serialhexi'll let you know tomorrow06:27
serialhexor later today :P06:27
@sonney2kthe script you had first should work - the one where you add '\n' won't06:27
shogun-buildbotbuild #155 of csharp_modular is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #150 of ruby_modular is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #152 of r_modular is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #153 of java_modular is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #153 of lua_modular is complete: Success [build successful]  Build details are at
-!- heiko [~heiko@] has joined #shogun12:42
-!- heiko [~heiko@] has quit [Ping timeout: 258 seconds]13:02
-!- heiko [~heiko@] has joined #shogun13:13
@sonney2kheiko, hi!13:52
heikosonney2k, hi13:52
heikodont worry I will finish the migration soon13:52
@sonney2khow is it going?13:52
heikofine here, what about you?13:53
@sonney2kok, so soon means hours or days?13:53
heikoIt is already working, I am trying out some more examples to eliminate some possible bugs13:53
heikoOne thing sucks: I forgot my Laptop power cable at home and my remaining time is only 30mins13:54
heikoI think I will send a pull request tomorrow13:54
@sonney2kit would be optimal if we could try to work on the migrations this evening13:54
heikothen we will have to add the migraion methods to each class that has a changed parameter13:54
@sonney2kyes - I would have wished to start that tonight if you are that far already13:54
@sonney2kwe have only 2 days before the intended release data so time is pressing a bit now13:55
heikoyes true13:55
heikommh, I have some urgent stuff to do this evening, I will move out of my flat on friday and nothing is done yet13:56
heikoLets do it like this13:56
@sonney2kso if you have an example that shows how to use migrations and your migrations don't break anything I would like to merge even sooner13:56
heikoI will go home now and work there offline13:57
heikothen I will put a "pull"-able  patch together13:57
heikoand come to university this evening to send it13:57
@sonney2kor internet cafe or so13:57
heikoyes, but its not far away from here13:57
heikoI think it should work pretty fine13:58
heikoat least for numerical stuff13:58
@sonney2kwould be great - we have everything running now
@sonney2k(all examples work)13:58
heikothe most important parts are the actual migrate methods, which have to be written by hand for each changed parameter13:58
@sonney2kand documentations is also updated
@sonney2kheiko, yeah but we didn't do *huge* changes13:59
@sonney2kmostly additions13:59
@sonney2kor sgvector replacement13:59
heikoyes, I am trying that out in the moment13:59
@sonney2kso if you have an example on how to ignore new things / sgvector stuff then we should be able to get the old test suite to work14:00
heikothe map stuff works and also recursively replacing parameters using the migrate method14:00
heikook, I will do that now and then send a patch this evening14:01
heikoI will be here working on it tomorrow before noon also14:01
heikoOk then, see you this evening14:01
@sonney2kso there is a chance we can do it14:01
-!- heiko [~heiko@] has quit [Ping timeout: 258 seconds]14:07
-!- alesis-novik [~alesis@] has joined #shogun16:52
-!- blackburn [~blackburn@] has joined #shogun17:38
CIA-3shogun: Soeren Sonnenburg master * race079d / (28 files): add further c# examples -
@sonney2kblackburn, heiko said that he will be back w/ a patch tonight :)17:52
blackburnsonney2k: nice17:53
@sonney2kin the evening already17:53
blackburnany other news?17:53
@sonney2kso we might give the transitions a shot17:53
@sonney2kmore c# examples17:53
@sonney2kfrom daniel17:53
blackburnI'm pretty busy today with all the moving to samara from my home17:53
@sonney2kwhen are you leaving?17:53
blackburnalready in samar17:54
blackburnso now unpacking aha17:54
@sonney2kgzip -d17:54
blackburnwill do something with examples or so this night17:55
@sonney2kblackburn, hopefully you have some time tonight - maybe we can get the first few migration related things to work17:55
@sonney2kif things run smoothly - then we are good17:55
@sonney2kblackburn, btw octave is broken on the buildbot (still) so tests related to it might be a bit unreliable...17:55
CIA-3shogun: Soeren Sonnenburg master * rfc63be5 / src/CONTRIBUTIONS : mention daniel in contributions -
blackburnsonney2k: why octave is broken?19:34
-!- blackburn [~blackburn@] has quit [Ping timeout: 258 seconds]19:41
-!- blackburn [~blackburn@] has joined #shogun19:43
@sonney2kdebian unstabe mess20:20
-!- sonney2k [] has left #shogun ["Leaving"]20:31
-!- sonney2k [] has joined #shogun20:31
-!- mode/#shogun [+o sonney2k] by ChanServ20:31
-!- heiko [~heiko@] has joined #shogun20:45
-!- blackburn [~blackburn@] has quit [Ping timeout: 258 seconds]21:12
heikosonney2k, I just made my last "test"-example work, I will put together an example with a SGVector transition now21:20
heikoargh found another error21:31
heikook scalars work now21:37
heikommh sgvectors still make problems21:37
heikoremeber that "treated in the same way as scalar variables" thing?21:38
-!- blackburn [~blackburn@] has joined #shogun21:43
CIA-3shogun: Sergey Lisitsyn master * ra4d19ca / (6 files): Added few examples descriptions -
@sonney2kheiko, I don't understand...21:51
heikoi got problems with these new container types21:52
heikoCT_SGVECTOR etc21:52
heikobut I just thought of a way how to handle it21:52
heikoit will do it for SGVector for now21:52
heikothen you can have a look before i do the other ones21:52
heikothis is getting more and more complicated21:52
heikobut at least: scalars work :)21:52
heikotype changes, name changes, no changes - everything21:53
blackburnI have bad connection, what's up here?21:53
blackburnsonney2k: any help I can do?21:53
-!- alesis-novik [~alesis@] has quit [Remote host closed the connection]22:09
-!- blackburn [~blackburn@] has quit [Read error: Operation timed out]22:21
heikosonney2k, sorry the sgvector migration leads to a bunch of new problems22:26
heikoi will not finish this today22:26
@sonney2k what problems?22:26
heikocopying Tparameter instances (with their data) is complicated with these types22:27
heikoevery time, an SGVector has to be allocated, along with its data22:27
heikoand this has to be done when migrating22:28
heikoI am currently changing the migration method to be able to use already allocated data22:28
heikosorry that this takes so long, but the stuff is even more complicated than the model selection stuff22:29
heikosonney2k, I gotta go now, my girlfriend is killing me already, we have to clear up our flat. I will be back tomorrow morning and continue22:30
@sonney2khmmhh seems like the problem is more w/ this SGVector stuff22:30
heikonormaly it is easy to create a new TParameter instance22:30
heikosimply copy the m_parameter memory22:30
@sonney2kif we 'just' had double* stuff it would work I guess22:30
heikobut with SGVector this is not possible like this22:31
@sonney2kbut can't you do the same thing with the SGVector stuff?22:31
@sonney2kyou just say a=b22:31
heikono, the m_parameter field contains a vector22:31
heikothis has to be copied22:31
heikobut also the ->vector field22:31
@sonney2kthe vector has to be copied too?22:31
@sonney2knot just the ptr?22:31
heikocurrently, yes22:32
heikobut i will change this22:32
heikoalso, a=b only works if the types are clear22:32
@sonney2kptr should be sufficient22:32
@sonney2ka=b is the same like memccpy in our case22:32
@sonney2kdo you copy the double* ptr's or do you copy memory for that too?22:33
heikomemory too22:33
heikobut i will change it22:33
@sonney2kok that definitely needs to be changed22:34
heikowhen a parameter does not change from one version to next one22:34
@sonney2kdoesn't make sense to copy memory around22:34
heikowork in progress...22:34
heikoit is there because of this:22:35
heikowhen a migration is performed, a type may change or something else may happen with the data22:35
heikoso that is cannot be used, but has to be converted somehow22:35
heikoin this case, migrate returns a new TParameter instance (with allocated memory) with the converted data22:35
heikoand this data has to be deleted afterwards (because of multiple migration steps)22:36
heikoi did this as a first step22:36
heikoand then noticed that sometimes, data does not get changed22:36
@sonney2kI would say one should assume that things that are migrated from have to stay constant22:36
heikoin this case, no copy has to be returned, but also this stuff must not be deleted afterwords22:36
@sonney2kand the things we migrate to might have to be clones22:37
heikoyes, thats how i want to do it22:37
heikobut just realised it22:37
@sonney2kyeah it seems really much more involved than I ever imagined22:37
heikoyes, its complicated stuff22:38
@sonney2kwhat is important now is that we get something to work good enough for the initial migration22:38
@sonney2kwe can polish this later on I would say22:38
@sonney2kcurrent migrations aren't too difficult - so it shouldn't be too hard22:38
heikoyes, but the infrastructure has to be capable of more complex migrations so that we avoid doing all this stuff all over again in the next release22:39
@sonney2kthat is true22:39
heikobut is has not to be implemented yet22:39
heikosonney2k, I will go for it tomorrow and hopefully it will go then, I have the feeling that "it will work soon" for quite some time now22:40
@sonney2kyeah that's not goo22:40
@sonney2kapologies to your gf22:40
heiko :) thanks22:41
heikoso cu22:41
-!- heiko [~heiko@] has quit [Ping timeout: 258 seconds]22:45
-!- blackburn [~blackburn@] has joined #shogun23:25
-!- blackburn1 [~blackburn@] has joined #shogun23:29
-!- blackburn [~blackburn@] has quit [Ping timeout: 245 seconds]23:29
blackburn1have read the logs23:58
blackburn1didn't understand what's up with serialization :)23:58
--- Log closed Tue Aug 30 00:00:37 2011