--- Log opened Wed Aug 29 00:00:17 2012 | ||
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 00:14 | |
-!- sr___ [u5548@gateway/web/irccloud.com/x-fxvqghaankjqwlrp] has joined #shogun | 01:35 | |
shogun-buildbot | build #80 of nightly_default is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/80 | 03:55 |
---|---|---|
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has joined #shogun | 08:43 | |
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has quit [Client Quit] | 08:47 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 08:48 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has joined #shogun | 09:39 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 09:44 | |
-!- av3ngr [~av3ngr@60-241-222-244.static.tpgi.com.au] has joined #shogun | 10:07 | |
-!- av3ngr [~av3ngr@60-241-222-244.static.tpgi.com.au] has left #shogun [] | 10:07 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 246 seconds] | 11:52 | |
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun | 12:06 | |
-!- fjig [~fjig@orange-10.csc.kth.se] has joined #shogun | 14:06 | |
-!- fjig is now known as n4nd0 | 14:06 | |
blackburn | n4nd0: new nickname? :) | 14:07 |
n4nd0 | blackburn: connected from university :) | 14:09 |
n4nd0 | irssi took my username in this network | 14:09 |
n4nd0 | blackburn: when are we planning to release btw? | 14:11 |
n4nd0 | do you think everything is prepared? | 14:11 |
blackburn | n4nd0: heiko made me happy about serialization | 14:12 |
blackburn | it is broken :D | 14:12 |
n4nd0 | fuck... really? | 14:12 |
n4nd0 | everything related to serialization? | 14:12 |
n4nd0 | it sounds like a big issue | 14:12 |
blackburn | I am unsure what exactly is broken | 14:12 |
blackburn | worst thing heiko can't take care of that | 14:13 |
n4nd0 | mmm I think I could try to look at it tonight | 14:15 |
n4nd0 | is there any test where it breaks? to have sth to start with | 14:15 |
blackburn | n4nd0: I asked heiko - lets wait for answer | 14:22 |
n4nd0 | ok | 14:24 |
blackburn | n4nd0: we have quite a few serialization examples in examples/libshogun | 14:34 |
blackburn | you could try to track errors with valgrind | 14:34 |
n4nd0 | ok | 14:34 |
blackburn | I am going to get home soon so I will try to track down errors tonight too | 14:48 |
n4nd0 | ok | 14:58 |
n4nd0 | I will see you there then | 14:58 |
n4nd0 | I am going to class now | 14:59 |
n4nd0 | bye | 14:59 |
-!- n4nd0 [~fjig@orange-10.csc.kth.se] has quit [Quit: leaving] | 14:59 | |
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [Quit: Page closed] | 15:08 | |
CIA-52 | shogun: Soeren Sonnenburg master * r5398c6e / (3 files in 3 dirs): fix up clustering evaluation example - http://git.io/_AL1pg | 16:30 |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun | 17:33 | |
-!- gsomix [~gsomix@178.45.88.247] has joined #shogun | 17:33 | |
gsomix | good evening | 17:33 |
n4nd0 | hi | 17:35 |
-!- gsomix [~gsomix@178.45.88.247] has quit [Remote host closed the connection] | 17:44 | |
@sonney2k | shogun-buildbot, force build nightly_default | 17:45 |
shogun-buildbot | build forced [ETA 37m15s] | 17:45 |
shogun-buildbot | I'll give a shout when the build finishes | 17:45 |
n4nd0 | hi sonney2k | 17:55 |
n4nd0 | did you read about this issue we were talking about yesterday related to using MKL with pnorm larger than one? | 17:55 |
-!- gsomix [~gsomix@178.45.88.247] has joined #shogun | 18:13 | |
shogun-buildbot | build #81 of nightly_default is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/81 | 18:18 |
-!- audy [~audy@unaffiliated/audy] has quit [Quit: ZNC - http://znc.sourceforge.net] | 18:58 | |
-!- audy [~audy@heyaudy.com] has joined #shogun | 19:00 | |
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving] | 19:08 | |
-!- blackburn [~blackburn@188.168.5.215] has joined #shogun | 19:13 | |
@sonney2k | blackburn, hey... what is wrong with the string kernels? why are they suddently marked unstable? | 19:14 |
blackburn | sonney2k: because they were and are failing tests | 19:14 |
@sonney2k | blackburn, the old regression tests? or which? | 19:14 |
blackburn | yes, old ones | 19:15 |
@sonney2k | did they fail back then too? | 19:15 |
@sonney2k | (like when we released 1.0 last year?) | 19:15 |
blackburn | for 1.0 and 1.1 yes | 19:15 |
@sonney2k | I can tell that noone can report anything | 19:16 |
@sonney2k | these are too involved | 19:16 |
blackburn | sonney2k: I am going through serialization stuff now - not really understand anything | 19:16 |
blackburn | sonney2k: yes and because of it I think they are unstable | 19:17 |
@sonney2k | blackburn, whats wrong with serializaton? | 19:17 |
blackburn | sonney2k: heiko says it is broken somehow | 19:17 |
@sonney2k | in which way? | 19:17 |
@sonney2k | what is not working? | 19:17 |
blackburn | I want to understand that too | 19:17 |
blackburn | well it is leaking | 19:17 |
@sonney2k | blackburn, do the 'new' regression tests work? | 19:17 |
@sonney2k | 'just' memory leaks? | 19:17 |
blackburn | sonney2k: yes seems to be that only leaks | 19:18 |
blackburn | sonney2k: unit tests are working yes | 19:18 |
@sonney2k | ahh ok then I know | 19:18 |
blackburn | sonney2k: is there a code where sgvector is being created? | 19:18 |
@sonney2k | it is because of SGVector and friends I think | 19:18 |
blackburn | on deserialization? | 19:18 |
@sonney2k | the reference count needs to be saved too IMHO | 19:18 |
blackburn | yes it should | 19:19 |
blackburn | I am confused with a lot of code there | 19:19 |
@sonney2k | blackburn, any other issues that need attention? | 19:20 |
blackburn | no, I don't think so | 19:20 |
blackburn | sonney2k: do you have time to check serialization? | 19:21 |
@sonney2k | blackburn, btw I learned that our 'regression' tests are called 'integration' tests | 19:21 |
@sonney2k | so I would rather rename them now | 19:21 |
@sonney2k | blackburn, no | 19:21 |
@sonney2k | that is too involved | 19:22 |
@sonney2k | I am afraid only heiko can fix things | 19:22 |
blackburn | heiko is out now | 19:22 |
blackburn | :) | 19:22 |
@sonney2k | blackburn, when is his deadline for handing in the masters thesis? | 19:22 |
blackburn | he is busy till 8th | 19:23 |
@sonney2k | + 2 days to recover | 19:23 |
@sonney2k | blackburn, I realize that we have quite some crappy examples under python_modular | 19:23 |
@sonney2k | not really functions | 19:23 |
@sonney2k | so not useful for integration tests | 19:24 |
@sonney2k | some even pull stuff of the web | 19:24 |
blackburn | that's all pretty bad | 19:24 |
@sonney2k | also some have documentation on top | 19:24 |
@sonney2k | instead of descriptions/ folder | 19:24 |
blackburn | sonney2k: are you back to your job already too? | 19:25 |
@sonney2k | no | 19:25 |
blackburn | when will you? | 19:25 |
@sonney2k | but next week | 19:25 |
blackburn | I see | 19:25 |
@sonney2k | will mean less time (can that be?) | 19:25 |
blackburn | less? | 19:25 |
@sonney2k | too bad I got sick now that I in principle would have had time | 19:25 |
@sonney2k | at least I can type again ;) | 19:26 |
blackburn | I don't know whether we could be ready | 19:26 |
@sonney2k | no idea either | 19:28 |
@sonney2k | if not then not | 19:28 |
blackburn | sonney2k: I like rc concept so may be we could release 2.0rc | 19:29 |
blackburn | :D | 19:29 |
blackburn | it is actually seems to be a good idea for me | 19:30 |
CIA-52 | shogun: Soeren Sonnenburg master * r550727e / (264 files in 30 dirs): rename regression test directory into integration tests to avoid - http://git.io/ICIggQ | 19:30 |
blackburn | because no way to release 2.0 without bugs we would have to fix in next 2-3 weeks | 19:30 |
@sonney2k | but why release RC then if we know the bugs :D | 19:31 |
@sonney2k | otherwise I am fine with rc's | 19:31 |
blackburn | sonney2k: no we don't know all bugs | 19:31 |
blackburn | I expect more anyway | 19:31 |
blackburn | sonney2k: uhhhhh ref count currently is pain in the ass | 19:33 |
blackburn | more or less I realized how that works | 19:33 |
@sonney2k | ? | 19:36 |
shogun-buildbot | build #381 of deb2 - static_interfaces is complete: Failure [failed test cmdline_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/381 blamelist: Soeren Sonnenburg <sonne@debian.org> | 19:38 |
blackburn | sonney2k: I don't know how to handle that at all | 19:40 |
blackburn | sgvector could be referenced from basically every other object | 19:41 |
CIA-52 | shogun: Soeren Sonnenburg master * rfeafbf7 / src/Makefile.template : - adjust regression -> integration in makefile - http://git.io/4dlw7Q | 19:43 |
@sonney2k | blackburn, yes | 19:43 |
blackburn | that means we can't determine actual refcount | 19:44 |
blackburn | actually after deserialization it seems to be 1 | 19:45 |
blackburn | for every field | 19:45 |
blackburn | I do not have any scenario with refcount=2 initially | 19:45 |
@sonney2k | blackburn, the only thing I can think of is that one has two RealFeatures objects sharing the same SGMatrix or so | 19:46 |
@sonney2k | ( or CLabels) | 19:46 |
@sonney2k | then current serialization would save things twice | 19:46 |
blackburn | yeah that's one | 19:46 |
blackburn | after deserialization it will become correct again | 19:47 |
blackburn | right? | 19:47 |
@sonney2k | blackburn, I think the proper fix would be to use gsomix' CMap class to store all the things that are to be mapped first | 19:47 |
@sonney2k | then later save them | 19:47 |
blackburn | ?? | 19:47 |
@sonney2k | this way it can be deduplicated | 19:47 |
@sonney2k | and refcounts can be computed | 19:47 |
blackburn | no I don't want to implement that now :D | 19:48 |
blackburn | lets assume it is 1 | 19:48 |
blackburn | after deserialization | 19:48 |
@sonney2k | I didnt' say that this should be done *now* | 19:48 |
@sonney2k | but I think that this is the only clean way | 19:48 |
@sonney2k | otherwise we save things twice or even more often | 19:49 |
@sonney2k | and of course behavior changes since objects that were just pointers to the same memory location | 19:49 |
@sonney2k | are now clones | 19:49 |
blackburn | sonney2k: okay I will add reset_refcount method and call it after deserialization | 19:51 |
blackburn | that should fix the process a little | 19:51 |
@sonney2k | blackburn, why reset? | 19:51 |
@sonney2k | 1 is OK | 19:51 |
blackburn | sonney2k: reset = set to 1 | 19:51 |
@sonney2k | blackburn, yeah but it is one no? | 19:51 |
blackburn | why? where is that set? | 19:52 |
@sonney2k | what is it? | 19:52 |
blackburn | okay you are creating machine | 19:52 |
blackburn | play around with w | 19:52 |
blackburn | w refcounter is 9 | 19:52 |
blackburn | then you load it | 19:52 |
blackburn | hmmmmmm | 19:53 |
blackburn | then no need to touch it?? | 19:53 |
blackburn | okay I lost the idea what I need to fix | 19:53 |
blackburn | :D | 19:53 |
@sonney2k | yeah I also don't know | 19:54 |
@sonney2k | it should not create leaks | 19:54 |
@sonney2k | or heiko means some other leaks | 19:54 |
@sonney2k | maybe *when* (de)serializing? | 19:55 |
@sonney2k | I mean in that code? | 19:55 |
blackburn | sonney2k: if you run any base or serialization or parameter libshogun example it leaks | 19:56 |
blackburn | sonney2k: okay so what about 2.0rc on 1st of September and 2.0 in the end of September? | 20:01 |
shogun-buildbot | build #483 of deb3 - modular_interfaces is complete: Failure [failed test python_modular] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/483 blamelist: Soeren Sonnenburg <sonne@debian.org> | 20:04 |
blackburn | wiking: please fix latent svm unit test | 20:05 |
shogun-buildbot | build #382 of deb2 - static_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/382 | 20:11 |
shogun-buildbot | build #484 of deb3 - modular_interfaces is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/484 | 20:43 |
wiking | blackburn: mmmm | 20:52 |
wiking | ok | 20:52 |
@sonney2k | wiking, what are the required packages to run the unit tests and how? | 21:05 |
@sonney2k | blackburn, do you also get an error in regression_gaussian_process_modular.py | 21:05 |
wiking | sonney2k: ubuntu? | 21:05 |
wiking | sonney2k: you need google-mock | 21:06 |
wiking | and need to install http://code.google.com/p/googletest/ as there's no package for it under ubuntu (just -dev package) | 21:06 |
@sonney2k | wiking, both debs? | 21:07 |
wiking | sonney2k: haven't checked | 21:07 |
wiking | maybe debian has it | 21:07 |
@sonney2k | libgtest-dev ? | 21:07 |
wiking | sonney2k: have u managed to get the static-build work? | 21:07 |
wiking | sonney2k: u would need libgtest0 | 21:07 |
wiking | http://packages.debian.org/squeeze/libgtest0 | 21:08 |
@sonney2k | doing | 21:08 |
wiking | so debian has it | 21:08 |
@sonney2k | (wheezy has it) | 21:08 |
@sonney2k | wiking, now what? | 21:08 |
wiking | ok after this | 21:08 |
wiking | run ./configure | 21:08 |
wiking | and hopefully it'll detect the mocking framework for you | 21:09 |
@sonney2k | nope | 21:09 |
wiking | mmm | 21:09 |
wiking | can u paste me here the configure.log part? | 21:09 |
@sonney2k | I installed google-mock | 21:09 |
@sonney2k | /usr/bin/ld: cannot find -lgmock | 21:09 |
@sonney2k | /usr/bin/ld: cannot find -lgtest | 21:09 |
wiking | yeah google-mock will install libgtest-dev | 21:09 |
wiking | but not libgtest0 | 21:10 |
wiking | can u send me the compiling code? | 21:10 |
wiking | what are the actual flags | 21:10 |
@sonney2k | wiking, I have /usr/lib/libgtest.so.0 | 21:11 |
@sonney2k | but no libgmock or so | 21:11 |
wiking | oooh | 21:11 |
wiking | dpkg -L google-mock | 21:11 |
wiking | does not install libgmock ?1 | 21:11 |
wiking | ?! | 21:11 |
@sonney2k | nope | 21:11 |
wiking | is this deb wheezy? | 21:12 |
@sonney2k | dpkg -L google-mock | grep .so | 21:12 |
@sonney2k | empty | 21:12 |
wiking | rotfl | 21:12 |
@sonney2k | y | 21:12 |
blackburn | sonney2k: I'll check but it was failing before, yes | 21:12 |
wiking | what does it have then? | 21:12 |
@sonney2k | gmock_gen and a buch of headers | 21:12 |
wiking | m,mmm | 21:13 |
wiking | does it have maybe then libgmock ? | 21:13 |
@sonney2k | wiking, | 21:13 |
@sonney2k | The Google C++ Mocking Framework uses conditional compilation for some | 21:13 |
@sonney2k | things. Because of the C++ "One Definition Rule", gmock must be | 21:13 |
@sonney2k | compiled with exactly the same flags as your C++ code under test. | 21:13 |
@sonney2k | Because this is hard to manage, upstream no longer recommends using | 21:13 |
@sonney2k | precompiled libraries. | 21:13 |
@sonney2k | so | 21:13 |
@sonney2k | the new recommendation is: | 21:14 |
@sonney2k | g++ -I/usr/src/gmock -c /usr/src/gmock/src/gmock-all.cc | 21:14 |
@sonney2k | g++ -I/usr/src/gtest -c /usr/src/gtest/src/gtest-all.cc | 21:14 |
@sonney2k | ar -rv libgmock.a gmock-all.o gtest-all.o | 21:14 |
@sonney2k | to build libgmock statically! | 21:14 |
wiking | lol | 21:14 |
@sonney2k | everyone has to do it before running tests | 21:14 |
@sonney2k | wiking, same with libgtest btw! | 21:15 |
wiking | yeah i gathered now | 21:15 |
wiking | that this is why there's no libgtest0 | 21:15 |
@sonney2k | exactly | 21:15 |
@sonney2k | actually squeeze still had libgtest | 21:16 |
@sonney2k | wheezy does not | 21:16 |
blackburn | sonney2k: yeah failing still | 21:30 |
blackburn | sonney2k: we are waiting for jacob's fix | 21:35 |
blackburn | okay leaving now | 21:35 |
-!- blackburn [~blackburn@188.168.5.215] has quit [Quit: Leaving.] | 21:36 | |
naywhayare | hello there, can someone tell me if I am performing this task in the most computationally efficient way? (task to follow) | 22:06 |
naywhayare | I have one dataset (no test/train split) and wish the find the K nearest neighbors (index of neighbors and distances) of each point in the dataset | 22:07 |
naywhayare | the only way I can seem to find to do this is to use the KNN classifier, which is more than I want to do; I just want the neighbors, no classification | 22:07 |
naywhayare | is there a way to do this? currently I am using the command-line interface but it seems to me like that has less functionality than some other interfaces to the library | 22:08 |
naywhayare | thanks in advance for any help. | 22:09 |
gsomix | sonney2k, hey. :) | 22:21 |
@sonney2k | naywhayare, I guess the most effictive would be to create the kd-tree like k-nn does yourself but that would mean you have to get your hands dirty too | 22:26 |
@sonney2k | and write C++ cod | 22:26 |
@sonney2k | e | 22:26 |
@sonney2k | gsomix, hey... | 22:26 |
@sonney2k | wiking, do you intend to fix the gtest/gmock stuff? | 22:27 |
gsomix | sonney2k, how can I add something in python_modular/tools? | 22:28 |
@sonney2k | just add it there... | 22:34 |
@sonney2k | but wnat do you need? | 22:34 |
naywhayare | sonney2k: I don't mind getting my hands dirty; if there's not a nice way to do it through one of the interfaces I'll just write up a little C++ utility | 22:34 |
naywhayare | thanks for th einfo | 22:34 |
naywhayare | *the info | 22:34 |
@sonney2k | naywhayare, we actually don't mind you do it and share the code later on via a PR such that it can be used from some other interface | 22:35 |
naywhayare | okay, I'll share the code but it'll be brutally simple and probably not useful to anybody but me | 22:36 |
gsomix | sonney2k, add in repo. | 22:36 |
gsomix | hm... I want classifier_multiclass_shared.py -> tools/multiclass_shared.py | 22:36 |
@sonney2k | gsomix, git mv classifier_multiclass_shared.py tools/multiclass_shared.py | 22:40 |
gsomix | sonney2k, ok | 22:40 |
* sonney2k Zzzzz | 22:40 | |
gsomix | sonney2k, nite | 22:41 |
naywhayare | am I correct in understanding that the CKNN class either uses cover trees or does the naive O(n^2) computation? I don't see kd-trees being created there | 22:53 |
naywhayare | forgive me if I've missed something simple as I am unfamiliar with shogun internals | 22:53 |
--- Log closed Thu Aug 30 00:00:17 2012 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!