--- Log opened Mon May 06 00:00:32 2013 | ||
shogun-buildbot | build #913 of deb2 - static_interfaces is complete: Failure [failed install test octave_static] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/913 blamelist: Soeren Sonnenburg <sonne@debian.org> | 00:31 |
---|---|---|
-!- travis-ci [~travis-ci@ec2-184-73-20-211.compute-1.amazonaws.com] has joined #shogun | 00:35 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg'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/6906494 | 00:35 |
-!- travis-ci [~travis-ci@ec2-184-73-20-211.compute-1.amazonaws.com] has left #shogun [] | 00:35 | |
van51 | may be a silly question, but how can I run the tests in a specific unit test file? | 01:03 |
shogun-buildbot | build #1055 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/1055 | 01:07 |
@lisitsyn | van51: --gtest-filter=Test.* | 01:14 |
@lisitsyn | like that | 01:14 |
van51 | lisitsyn: don't know where that goes :p | 01:20 |
@lisitsyn | van51: as a key of test binary | 01:20 |
van51 | i wrote a main function inside one :$ | 01:20 |
@lisitsyn | van51: main? why? | 01:21 |
van51 | well, I wanted to run the TEST() in a file | 01:21 |
van51 | so I added a main and called RUN_ALL_TESTS inside that file | 01:21 |
van51 | noobish way, but I haven't used a testing framework without an IDE before :/ | 01:23 |
@lisitsyn | I see | 01:25 |
@lisitsyn | not sure about IDE usage here | 01:25 |
van51 | lisitsyn: can you explain it though a bit? | 01:27 |
van51 | all I know is to call make tests | 01:27 |
van51 | but I have a few (not mine) failing | 01:27 |
@lisitsyn | van51: alright let me check | 01:27 |
@lisitsyn | oops I broke gmock here | 01:30 |
van51 | I wanted to ask about that too :p | 01:31 |
van51 | my ./configure didn't spot it | 01:31 |
@lisitsyn | van51: os? | 01:32 |
van51 | ubuntu | 01:32 |
@lisitsyn | van51: I am afraid manual install is the way | 01:32 |
van51 | ah ok | 01:33 |
van51 | lisitsyn: suppose I do that, instead of make tests, how do I select the specific test? | 01:33 |
@lisitsyn | van51: in a minute | 01:35 |
van51 | lisitsyn: sure, do your job :) | 01:36 |
@lisitsyn | van51: ./shogun-unit-test --gtest_filter=LaplacianInferenceMethod.get_alpha | 01:39 |
@lisitsyn | as an example | 01:39 |
@lisitsyn | have to sleep now cu | 01:40 |
van51 | lisitsyn: thank you | 01:40 |
van51 | goodnight | 01:40 |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has joined #shogun | 02:32 | |
-!- van51 [~van51@athedsl-318776.home.otenet.gr] has left #shogun ["PING 1367803522"] | 03:25 | |
shogun-buildbot | build #383 of nightly_default is complete: Failure [failed install test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/383 | 04:25 |
-!- nube [~rho@116.90.239.3] has joined #shogun | 05:20 | |
-!- lisitsyn [~blackburn@109-226-120-160.clients.tlt.100megabit.ru] has left #shogun [] | 06:37 | |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 256 seconds] | 08:11 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 08:13 | |
@sonney2k | shogun-buildbot, force build 'nightly_default' | 08:21 |
shogun-buildbot | build forced [ETA 35m39s] | 08:21 |
shogun-buildbot | I'll give a shout when the build finishes | 08:21 |
@sonney2k | shogun-buildbot, force build 'deb2 - static_interfaces' | 08:22 |
shogun-buildbot | build #914 forced | 08:22 |
shogun-buildbot | I'll give a shout when the build finishes | 08:22 |
-!- iglesiasg [~nando@88.128.80.7] has joined #shogun | 08:29 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 08:29 | |
-!- iglesiasg [~nando@88.128.80.7] has quit [Quit: leaving] | 08:34 | |
shogun-buildbot | build #914 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/914 | 08:34 |
shogun-buildbot | build #384 of nightly_default is complete: Failure [failed install test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/384 | 08:50 |
-!- sijin [~smuxi@144.214.222.109] has joined #shogun | 09:09 | |
-!- erlenda [~erlenda@fw-oslo.intelcom.no] has joined #shogun | 09:15 | |
@sonney2k | shogun-buildbot, force build 'nightly_default' | 09:36 |
shogun-buildbot | build forced [ETA 35m39s] | 09:36 |
shogun-buildbot | I'll give a shout when the build finishes | 09:36 |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 272 seconds] | 09:37 | |
shogun-buildbot | build #385 of nightly_default is complete: Failure [failed install test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/385 | 10:01 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 10:03 | |
-!- mode/#shogun [+o lisitsyn] by ChanServ | 10:03 | |
@sonney2k | shogun-buildbot, force build 'nightly_default' | 10:09 |
shogun-buildbot | build forced [ETA 35m39s] | 10:09 |
shogun-buildbot | I'll give a shout when the build finishes | 10:09 |
@lisitsyn | sonney2k: how do you feel? | 10:10 |
@sonney2k | sick | 10:12 |
@lisitsyn | badz | 10:16 |
-!- ozan [~ozan@ee-167-078.eee.metu.edu.tr] has joined #shogun | 10:19 | |
ozan | before starting to write unit tests I would like to solve build fail for QDA | 10:23 |
ozan | I got an error in Travis | 10:23 |
ozan | https://travis-ci.org/shogun-toolbox/shogun/jobs/6859829 | 10:23 |
ozan | I think issue is related to python example | 10:23 |
ozan | however it is not included in the PR | 10:24 |
@lisitsyn | sonney2k: my flight is finally booked | 10:27 |
@lisitsyn | so if I am not hit by bus I am in berlin 9-15 | 10:27 |
@lisitsyn | ozan: yeah we have a trouble with it for some reason | 10:29 |
ozan | so what should I do then | 10:32 |
@lisitsyn | ozan: well someone of us has to fix it at some point | 10:36 |
@lisitsyn | why do you want to solve it for QDA? | 10:36 |
ozan | I just wanted to get a successful build message | 10:37 |
ozan | if not necessary I will proceed with unit tests then | 10:37 |
shogun-buildbot | build #756 of cyg1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/756 blamelist: Soeren Sonnenburg <sonne@debian.org> | 10:42 |
shogun-buildbot | build #386 of nightly_default is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/nightly_default/builds/386 | 10:44 |
ozan | is there any sample unit test for the functionality similar to QDA which I can check ? | 10:45 |
@lisitsyn | ozan: oh sorry I didn't realize it is qda example that fails | 10:46 |
@lisitsyn | that's worse! any idea why it produces different result? | 10:46 |
-!- Yanglittle [deb20afa@gateway/web/freenode/ip.222.178.10.250] has joined #shogun | 10:48 | |
ozan | I have an idea | 10:48 |
ozan | at some point | 10:48 |
ozan | I print labels for debug | 10:48 |
Yanglittle | excuse me, didn't the clustering deal with sparse data? | 10:48 |
ozan | but travis fails at that line | 10:49 |
ozan | however I did not included example file in my commit | 10:49 |
-!- sijin [~smuxi@144.214.222.109] has quit [Ping timeout: 264 seconds] | 10:49 | |
Yanglittle | excuse me, didn't the clustering deal with sparse data? | 10:49 |
-!- travis-ci [~travis-ci@ec2-184-73-20-211.compute-1.amazonaws.com] has joined #shogun | 10:49 | |
travis-ci | [travis-ci] it's Soeren Sonnenburg'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/6916233 | 10:49 |
-!- travis-ci [~travis-ci@ec2-184-73-20-211.compute-1.amazonaws.com] has left #shogun [] | 10:49 | |
Yanglittle | excuse me, didn't the clustering deal with sparse data? | 10:53 |
@lisitsyn | Yanglittle: k-means? | 10:55 |
Yanglittle | yes. | 10:55 |
@lisitsyn | should work with sparse distance but I am afraid it cluster centers would be dense | 10:55 |
ozan | Can I use travis without making a pull request ? | 11:24 |
-!- nube [~rho@116.90.239.3] has joined #shogun | 11:28 | |
@lisitsyn | ozan: yeah you may set up travis for your fork | 11:28 |
ozan | thanks, I will set-up travis and try to solve the issue | 11:35 |
-!- nube [~rho@116.90.239.3] has quit [Ping timeout: 252 seconds] | 11:37 | |
-!- Yanglittle [deb20afa@gateway/web/freenode/ip.222.178.10.250] has quit [Quit: Page closed] | 11:44 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 11:49 | |
-!- travis-ci [~travis-ci@ec2-184-73-20-211.compute-1.amazonaws.com] has joined #shogun | 11:53 | |
travis-ci | [travis-ci] it's ozansener's turn to pay the next round of drinks for the massacre he caused in ozansener/shogun: http://travis-ci.org/ozansener/shogun/builds/6917797 | 11:53 |
-!- travis-ci [~travis-ci@ec2-184-73-20-211.compute-1.amazonaws.com] has left #shogun [] | 11:53 | |
-!- travis-ci [~travis-ci@ec2-184-73-30-241.compute-1.amazonaws.com] has joined #shogun | 12:11 | |
travis-ci | [travis-ci] it's ozansener's turn to pay the next round of drinks for the massacre he caused in ozansener/shogun: http://travis-ci.org/ozansener/shogun/builds/6917823 | 12:11 |
-!- travis-ci [~travis-ci@ec2-184-73-30-241.compute-1.amazonaws.com] has left #shogun [] | 12:11 | |
-!- nube [~rho@116.90.239.3] has quit [Quit: Leaving.] | 12:19 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 12:20 | |
-!- abinayam [3d0c1363@gateway/web/freenode/ip.61.12.19.99] has joined #shogun | 12:41 | |
-!- abinayam [3d0c1363@gateway/web/freenode/ip.61.12.19.99] has quit [Client Quit] | 12:43 | |
@wiking | ozan: plz remove the announcement from your .travis.yml conf... | 13:39 |
ozan | wiking: I removed the announcement sorry for the mess | 13:56 |
-!- nube [~rho@116.90.239.3] has quit [Read error: Operation timed out] | 14:06 | |
-!- nube [~rho@116.90.239.3] has joined #shogun | 14:06 | |
-!- nube [~rho@116.90.239.3] has quit [Client Quit] | 14:07 | |
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun | 15:25 | |
votjakovr | Hi everyone! I encountered a problem. When I run integration tests on my local machine, every test is OK, but travis-ci fails on two tests. Any ideas why does it happen? | 15:35 |
@lisitsyn | votjakovr: that's kind of mystery yeah | 15:36 |
votjakovr | lisitsyn: Yep. Most interesting, that I generated these two tests. | 15:39 |
@lisitsyn | votjakovr: some platform-dependent stuff - hell to debug | 15:41 |
votjakovr | lisitsyn: And I tried different configurations on my machine, but they are still OK | 15:43 |
-!- erlenda [~erlenda@fw-oslo.intelcom.no] has quit [Remote host closed the connection] | 16:20 | |
-!- ozan [~ozan@ee-167-078.eee.metu.edu.tr] has quit [Quit: Leaving] | 16:22 | |
-!- nube [~rho@49.244.114.202] has joined #shogun | 16:28 | |
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds] | 16:38 | |
-!- sijin [~smuxi@144.214.222.109] has joined #shogun | 16:44 | |
@wiking | sonney2k: ping | 16:52 |
@sonney2k | wiking, pong | 18:35 |
@sonney2k | wiking, lisitsyn anything suspicious after the wheezy upgrade? | 18:35 |
-!- van51 [~van51@athedsl-318776.home.otenet.gr] has joined #shogun | 18:38 | |
@wiking | sonney2k: haven't checked.... question: i want to have a random unittest, but for some of the cases i would need to save a vector, which size if 320+. | 18:41 |
@wiking | sonney2k: maybe i should just test that 2 times running PRNG will produce the same random vector? | 18:41 |
@wiking | and maybe create an integration test for this | 18:42 |
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has joined #shogun | 18:42 | |
-!- pickle27 [~kevin@rcv3-lab-pc.ee.queensu.ca] has joined #shogun | 18:48 | |
-!- sumit [73f91219@gateway/web/freenode/ip.115.249.18.25] has quit [Ping timeout: 245 seconds] | 18:55 | |
@sonney2k | wiking, it makes totally sense to also store such a vector | 19:04 |
@sonney2k | wiking, you could either generate the vector from some python_modular code and then return it | 19:05 |
@sonney2k | then it is serialized and 1:1 compared | 19:05 |
@sonney2k | or write it somewhere to shogun/data/ | 19:05 |
-!- iglesiasg [d58f322f@gateway/web/freenode/ip.213.143.50.47] has joined #shogun | 19:33 | |
-!- mode/#shogun [+o iglesiasg] by ChanServ | 19:34 | |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 19:54 | |
shogun-notifier- | shogun: Kevin Hughes :develop * 2ab1521 / / (4 files): https://github.com/shogun-toolbox/shogun/commit/2ab152121f8296ccc2310902ab4ea0bd7a946454 | 19:54 |
shogun-notifier- | shogun: multiclass lda initial commit, algorithm is working. ouptut is confirmed to match scikit learn, I would assume that more work is required to integrate this new class into shogun fully | 19:54 |
shogun-notifier- | shogun: Kevin Hughes :develop * 2fd7892 / src/shogun/multiclass/MCLDA.cpp,src/shogun/multiclass/MCLDA.h: https://github.com/shogun-toolbox/shogun/commit/2fd789284a5a78ddefbcbe9739f4d7987ee60ba3 | 19:54 |
shogun-notifier- | shogun: MCLDA fixed the bug that caused a memory error on more complicated data | 19:54 |
shogun-notifier- | shogun: Kevin Hughes :develop * cdb25fe / src/shogun/multiclass/MCLDA.h: https://github.com/shogun-toolbox/shogun/commit/cdb25fe9da7fff1e03012393aa4a77223fa6715d | 19:54 |
shogun-notifier- | shogun: Update MCLDA.h | 19:54 |
shogun-notifier- | shogun: | 19:54 |
shogun-notifier- | shogun: fixed the tabs in MCLDA.h | 19:54 |
shogun-notifier- | shogun: Kevin Hughes :develop * ca4051a / src/shogun/multiclass/MCLDA.cpp: https://github.com/shogun-toolbox/shogun/commit/ca4051a1047a9284762742ca05501c80bee12b5e | 19:54 |
shogun-notifier- | shogun: Update MCLDA.cpp | 19:54 |
shogun-notifier- | shogun: | 19:54 |
shogun-notifier- | shogun: fixed tabbing in MCLDA.cpp | 19:54 |
shogun-notifier- | shogun: Kevin Hughes :develop * 9bdbff3 / src/shogun/multiclass/MCLDA.cpp: https://github.com/shogun-toolbox/shogun/commit/9bdbff3392e5d4f2c05c13ba83c4e4e862d53923 | 19:54 |
-!- lisitsyn1 [~blackburn@83.234.54.230] has joined #shogun | 19:54 | |
shogun-notifier- | shogun: Update MCLDA.cpp | 19:54 |
@iglesiasg | sonney2k: finally we got multiclass LDA | 19:55 |
@iglesiasg | I remember you wanted this last year | 19:55 |
lisitsyn1 | ahh I forgot to log out at job | 19:55 |
lisitsyn1 | :D | 19:55 |
lisitsyn1 | iglesiasg: spain? | 19:55 |
@iglesiasg | lisitsyn1: yes | 19:56 |
lisitsyn1 | pickle27: if you have any idea for an unit test for mc lda it would be very appreciated | 19:56 |
@sonney2k | iglesiasg, yeah from you right? | 19:56 |
@iglesiasg | I was in Frankfurt this morning, I trained my German for the workshop :D | 19:56 |
@sonney2k | lisitsyn1, just authorize to nickserv and own your other me | 19:57 |
lisitsyn1 | iglesiasg: oh I am doing quite well with it! I now know 4 words | 19:57 |
@iglesiasg | sonney2k: no, pickle27 has done it | 19:57 |
lisitsyn1 | tegel, schillstrasse | 19:57 |
lisitsyn1 | kurfurstenstrasse | 19:57 |
@sonney2k | lisitsyn1, btw excellent that you merge things... I think we are a bit slow currently | 19:57 |
@sonney2k | iglesiasg, no I meant I asked you to do it last year :D | 19:57 |
lisitsyn1 | and schoenefeld! | 19:57 |
lisitsyn1 | :D | 19:57 |
lisitsyn1 | sonney2k: is KaDeWe that nice? | 19:57 |
@iglesiasg | sonney2k: oh yeah, but got my hands into the cover tree at the end | 19:58 |
@iglesiasg | haha I remember me struggling with it | 19:58 |
lisitsyn1 | JL tends to write one-shot code | 19:59 |
lisitsyn1 | once done impossible to modify | 19:59 |
pickle27 | I'll take a look at what scikit-learn has for a unit test and see if I can use their approach | 19:59 |
lisitsyn1 | pickle27: thanks! | 19:59 |
@iglesiasg | pickle27: and it would be sooo nice with eigen :) | 20:00 |
pickle27 | yeah I agree, Eigen code is much nicer to read | 20:00 |
@iglesiasg | lisitsyn1: do you think a version of MQDA/MLDA relying on eigen should be faster as well? | 20:00 |
pickle27 | I am still planning to get to QDA and LDA with Eigen in the next bit | 20:00 |
lisitsyn1 | I would not expect much faster | 20:01 |
@iglesiasg | pickle27: very nice, thank you! | 20:01 |
pickle27 | I don't know much about the speed difference between those libs but it will deffs be easier to follow, the whole row vs col pref is confusing | 20:02 |
shogun-buildbot | build #921 of bsd1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/921 blamelist: Kevin Hughes <kevin@kevin-M3920.(none)> | 20:02 |
@iglesiasg | pickle27: idd | 20:03 |
lisitsyn1 | pickle27: I didn't get the point - what's about row vs col? | 20:03 |
@iglesiasg | pickle27, lisitsyn1 : my guess is that getting rid of loops like https://github.com/shogun-toolbox/shogun/blob/master/src/shogun/multiclass/QDA.cpp#L232 with Eigen's vectorized operations should be faster | 20:03 |
shogun-buildbot | build #922 of bsd1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/922 blamelist: Kevin Hughes <kevinhughes27@gmail.com> | 20:04 |
pickle27 | fortran is column major and I believe Eigen3 is row major | 20:04 |
pickle27 | the input to the classifier is a matrix of samples as rows so it makes more sense for everything to follow with the row major mindset | 20:05 |
pickle27 | also these build failures aren't my fault right? it doesn't look related | 20:06 |
lisitsyn1 | pickle27: http://eigen.tuxfamily.org/dox/TopicStorageOrders.html | 20:06 |
lisitsyn1 | pickle27: eigen is not roww major by default | 20:06 |
lisitsyn1 | pickle27: yeah not your fault | 20:06 |
shogun-buildbot | build #468 of rpm1 - libshogun is complete: Failure [failed test] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/468 blamelist: Kevin Hughes <kevin@kevin-M3920.(none)> | 20:06 |
lisitsyn1 | uh | 20:07 |
lisitsyn1 | pickle27: this one is for you | 20:07 |
@iglesiasg | pickle27: but these warnings are | 20:07 |
@iglesiasg | http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/468/steps/compile/logs/warnings%20%282%29 | 20:07 |
@iglesiasg | :P | 20:07 |
@iglesiasg | aham and the last crash in classifier_lda? | 20:08 |
pickle27 | looking into the warnings - whats the crash? | 20:09 |
shogun-buildbot | build #916 of deb2 - static_interfaces is complete: Failure [failed test libshogun] Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb2%20-%20static_interfaces/builds/916 blamelist: Kevin Hughes <kevin@kevin-M3920.(none)> | 20:10 |
lisitsyn1 | sonney2k: you talkative! | 20:10 |
lisitsyn1 | :D | 20:10 |
shogun-buildbot | build #923 of bsd1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/923 | 20:10 |
shogun-buildbot | build #924 of bsd1 - libshogun is complete: Failure [failed configure] Build details are at http://www.shogun-toolbox.org/buildbot/builders/bsd1%20-%20libshogun/builds/924 blamelist: Kevin Hughes <kevin@kevin-M3920.(none)> | 20:10 |
lisitsyn1 | pickle27: I'd suggest to relax for a while now - it should build all the commits | 20:10 |
shogun-buildbot | build #469 of rpm1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/rpm1%20-%20libshogun/builds/469 | 20:10 |
lisitsyn1 | with different luck | 20:11 |
@sonney2k | pickle27, next time please submit a *single* commit | 20:11 |
@sonney2k | lisitsyn, what do you mean? | 20:11 |
pickle27 | yeah sorry about that | 20:11 |
lisitsyn1 | sonney2k: mentor stuff | 20:11 |
@sonney2k | ohh that email to mentors? | 20:11 |
@sonney2k | I send that 3 days ago... | 20:11 |
lisitsyn1 | hahah | 20:11 |
lisitsyn1 | I just received | 20:11 |
@sonney2k | me too | 20:11 |
@sonney2k | interesting | 20:11 |
lisitsyn1 | sonney2k: are you using russian post or what? | 20:11 |
@sonney2k | deutsche post is famous for speed! | 20:12 |
lisitsyn1 | sonney2k: I once ordered an replacement screen for my notebook and it was on the way for 2.5 month | 20:12 |
lisitsyn1 | beat it | 20:12 |
lisitsyn1 | from UK | 20:12 |
@iglesiasg | pickle27: comparing qda example https://github.com/shogun-toolbox/shogun/blob/master/examples/undocumented/libshogun/classifier_qda.cpp with yours | 20:13 |
@sonney2k | pickle27, well next time - you can just do git reset --soft <id> | 20:13 |
@iglesiasg | pickle27: I think it might be the missing SG_REF in output_labels | 20:13 |
@iglesiasg | pickle27: just output in your code, sorry | 20:13 |
@sonney2k | still you only need to SG_REF stuff if you pass it into other functions | 20:14 |
@sonney2k | not when you call obj->train() | 20:14 |
@iglesiasg | sonney2k: yeah, SG_REF and SG_UNREF for out_labels here looks unnecessary, doesn't it? | 20:15 |
pickle27 | yeah that makes sense, it was in the qda example so I left it in | 20:15 |
@iglesiasg | https://github.com/shogun-toolbox/shogun/blob/master/examples/undocumented/libshogun/classifier_qda.cpp | 20:15 |
shogun-buildbot | build #917 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/917 | 20:16 |
pickle27 | jumping back a bit to the Eigen discussion - I'm assuming that the rest of shogun uses Eigen as col major then? | 20:17 |
pickle27 | So I shall as well when I work on LDA and QDA with Eigen | 20:18 |
lisitsyn1 | pickle27: it is colmajor by default | 20:18 |
shogun-buildbot | build #757 of cyg1 - libshogun is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cyg1%20-%20libshogun/builds/757 | 20:18 |
pickle27 | so shogun uses it in the default way | 20:18 |
pickle27 | got it | 20:18 |
lisitsyn1 | yes sure | 20:18 |
@iglesiasg | colmajor is like an unspoken standard, isn't it? | 20:18 |
pickle27 | I don't know I think scikit learn is row major | 20:19 |
lisitsyn1 | iglesiasg: btw did you notice scikits treats feature matrices transposed? | 20:19 |
lisitsyn1 | when we have X they use X' | 20:19 |
@iglesiasg | pickle27: true, I have used some code derived from scikit which was row major | 20:20 |
pickle27 | I have a computer vision background so I think in row major | 20:20 |
@iglesiasg | lisitsyn1: yeah I noticed some things are transposed wrt to us in there | 20:21 |
@sonney2k | pickle27, iglesiasg yeah, matlab, octave, R all use column major | 20:23 |
@sonney2k | so that is where we come from | 20:23 |
pickle27 | makes sense | 20:23 |
@sonney2k | python is using c-style a[i][j] | 20:23 |
@sonney2k | I meant they wanted to be compatible with that | 20:23 |
@sonney2k | so numpy uses C-order | 20:23 |
@sonney2k | but can be tuned to work with fortran order | 20:24 |
pickle27 | yeah I have used that flag when porting matlab to numpy | 20:24 |
pickle27 | very hand :) | 20:24 |
pickle27 | handy* | 20:24 |
pickle27 | iglesiasg: btw I checked and that unused variable can be deleted but the other warning is okay its for the cov matrix which is stored but not used | 20:26 |
pickle27 | iglesiasg: should I make yet another (sorry) commit to fix this? | 20:27 |
@iglesiasg | pickle27: I have nothing against adding commits -- for the that's they way things work hehe | 20:28 |
@iglesiasg | pickle27: we should get rid of the warnings though | 20:29 |
@iglesiasg | pickle27: IIRC multiclass QDA does something similar as well, does the warning appear there too? | 20:30 |
-!- travis-ci [~travis-ci@ec2-107-21-82-250.compute-1.amazonaws.com] has joined #shogun | 20:30 | |
travis-ci | [travis-ci] it's Sergey Lisitsyn'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/6930799 | 20:30 |
-!- travis-ci [~travis-ci@ec2-107-21-82-250.compute-1.amazonaws.com] has left #shogun [] | 20:30 | |
@iglesiasg | pickle27: in case not (which I'd bet for) then there should be way to avoid the warning in lda too | 20:30 |
pickle27 | iglesiasg: okay I'll take a look | 20:34 |
-!- van51 [~van51@athedsl-318776.home.otenet.gr] has left #shogun ["PING 1367865330"] | 20:35 | |
pickle27 | iglesiasg: ah found it | 20:38 |
pickle27 | iglesiasg: qda has a method to return the cov matrix which I am missing | 20:38 |
@iglesiasg | pickle27: there you go | 20:42 |
@iglesiasg | pickle27: just add these two fixes for the warnings and for the test and make a new PR | 20:42 |
@iglesiasg | this should be much shorter and we should be able to merge it rather soon | 20:42 |
@iglesiasg | pickle27: regarding the test, could you please valgrind it to assure everything is all right? | 20:43 |
pickle27 | I'm looking into what happened now | 20:46 |
@iglesiasg | pickle27: I think it is that SG_UNREF(output) without previous SG_REF | 20:46 |
pickle27 | iglesiasg: there is a previous SG_REF | 20:47 |
pickle27 | iglesiasg: its right below the definition and right before the print statement | 20:47 |
@iglesiasg | pickle27: oh sorry, let me see again | 20:53 |
@iglesiasg | I am bling | 20:55 |
@iglesiasg | blind* | 20:55 |
gsomix | sonney2k, hello. there? | 20:57 |
@sonney2k | gsomix, yeah | 21:04 |
gsomix | sonney2k, can you explain a little your idea for "add factory classes for" issue? | 21:05 |
pickle27 | iglesiasg: I just rebuilt and ran the example on my machine and it went fine - what did you want me to do with valgrind I've never used that tool before | 21:05 |
@iglesiasg | pickle27: take a look to the Makefile | 21:06 |
@iglesiasg | or let me show you, one second | 21:06 |
@sonney2k | gsomix, I was thinking that it might be better to bundle all the obtain_from* stuff not in the respective e.g. label type but in some factory class e.g. CLabelFactory where all the static functions for converting labels are contained | 21:07 |
@iglesiasg | pickle27: compile the example with the -g for debuggin info | 21:08 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has quit [Read error: Connection reset by peer] | 21:08 | |
@iglesiasg | and do valgrind --tool=memcheck ./classifier_lda | 21:08 |
@iglesiasg | what do you see? | 21:08 |
-!- lisitsyn [~lisitsin@mxs.kg.ru] has joined #shogun | 21:09 | |
gsomix | sonney2k, hm, but what kind of methods? something like get_BinaryLabels_from_generic()? | 21:09 |
gsomix | BinaryLabels* get_BinaryLabels_from_generic(Labels*) | 21:09 |
gsomix | as example | 21:10 |
@iglesiasg | pickle27: or even better, valgrind --show-reachable=yes --leak-check=full ./classifier_lda | 21:10 |
pickle27 | iglesiasg: I do get an error, just trying to make sense of it | 21:11 |
@iglesiasg | pickle27: paste the output in pastebin if you want, I can take a look too | 21:12 |
@iglesiasg | the more eyes, the better finding errors | 21:12 |
-!- votjakovr [~votjakovr@host-46-241-3-209.bbcustomer.zsttk.net] has quit [Quit: Leaving] | 21:12 | |
pickle27 | iglesiasg: http://pastebin.com/ZafWLLJ0 | 21:12 |
@iglesiasg | pickle27: did you compile with -g? | 21:12 |
pickle27 | iglesiasg: I built using this g++ -g -o lda classifier_lda.cpp -lshogun | 21:13 |
pickle27 | oh wait | 21:13 |
pickle27 | then I ran the run thing woops | 21:13 |
@iglesiasg | what? | 21:14 |
pickle27 | I build a new target but ran valgrind on the old one, one sec | 21:14 |
@iglesiasg | ok | 21:14 |
pickle27 | iglesiasg: I updated the paste, looks error free? | 21:15 |
pickle27 | iglesiasg: or maybe not | 21:16 |
pickle27 | iglesiasg: yeah it still looks like it isn't running | 21:17 |
@iglesiasg | I see the same thing I think | 21:18 |
@iglesiasg | is it the same link? | 21:18 |
pickle27 | yeah I updated it though, I did build with -g | 21:19 |
pickle27 | but it doesn't seem to be 'attaching' | 21:19 |
@iglesiasg | nah just open a new pastebin :) | 21:19 |
@iglesiasg | I have never tried updating a already published one | 21:20 |
@iglesiasg | anyway, it is cheap to open a new one :D | 21:20 |
gsomix | sonney2k, I cannot figured out a more elegant solution that methods in specialized classes. | 21:20 |
pickle27 | iglesiasg: http://pastebin.com/TAWZC649 | 21:20 |
@iglesiasg | pickle27: this doesn't look normal | 21:21 |
gsomix | sonney2k, only automagically cast in swig typemaps. | 21:21 |
@iglesiasg | vex amd64->IR: unhandled instruction bytes: 0xC5 0xF9 0x57 0xC0 0xE8 0x73 0x4 0xFF | 21:21 |
@iglesiasg | pickle27: it looks like it doesn't compile correctly for your platform | 21:22 |
pickle27 | valgrind or the example? | 21:22 |
@iglesiasg | pickle27: you got to run this test without trouble before, didn't you? | 21:22 |
* gsomix afk | 21:22 | |
pickle27 | iglesiasg: yeah I can run the example fine right now | 21:22 |
@iglesiasg | pickle27: aham, pretty funny indeed | 21:23 |
@iglesiasg | let me see your valgrind line | 21:23 |
pickle27 | iglesiasg: valgrind --show-reachable=yes --leak-check=full ./lda | 21:23 |
@iglesiasg | and ./lda works fine? | 21:24 |
pickle27 | I just installed valgrind from apt | 21:24 |
pickle27 | yes | 21:24 |
@iglesiasg | run it again with -v please | 21:25 |
@iglesiasg | I don't think it will give us new info though | 21:25 |
pickle27 | iglesiasg: http://pastebin.com/ScJtjNxr | 21:25 |
@iglesiasg | pickle27: I don't get it | 21:26 |
pickle27 | iglesiasg: looks like it can't get the symbols from shogun and the other libs | 21:27 |
@iglesiasg | yeah | 21:27 |
@iglesiasg | the strange thing is that it pops up with valgrind | 21:27 |
@iglesiasg | pickle27: I am off for dinner now, will be back a bitlater | 21:28 |
pickle27 | okay later | 21:28 |
@sonney2k | I mean like CLabelFactory::to_binary(CLabels* labels); | 21:41 |
gsomix | sonney2k, ok, got it. | 21:48 |
gsomix | sonney2k, ahhh, I fell into the "overthinking" trap. :) | 21:54 |
@sonney2k | pickle27, compile shogun without optimizations (--disable-optimizations) | 22:11 |
@iglesiasg | re | 22:27 |
@iglesiasg | pickle27: did it get better after removing optimizations? | 22:27 |
@sonney2k | lisitsyn, lisitsyn1 - pretty interesting - for others the submissions curve is basically linear | 22:39 |
@sonney2k | and I just checked 2011,2012,2013 it is pretty exponential for us | 22:40 |
@sonney2k | ohh and we had 'only' 65 submissions in the first year | 22:41 |
@iglesiasg | ok, good night guys | 22:48 |
-!- iglesiasg [d58f322f@gateway/web/freenode/ip.213.143.50.47] has quit [Quit: Page closed] | 22:48 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 22:54 | |
pickle27 | I haven't had a chance to check yet | 22:56 |
pickle27 | I'll try and get to this later today | 22:56 |
-!- nube [~rho@49.244.114.202] has quit [Ping timeout: 256 seconds] | 23:50 | |
--- Log closed Tue May 07 00:00:34 2013 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!