Open in new window / Try shogun cloud
--- Log opened Fri Aug 31 00:00:17 2012
-!- gsomix [~gsomix@178.45.74.12] has quit [Ping timeout: 245 seconds]00:43
-!- Netsplit *.net <-> *.split quits: blackburn, yoh_, audy, shogun-buildbot, @sonney2k, zxtx, sr___, naywhayare, wiking, CIA-5202:46
-!- Netsplit over, joins: blackburn, zxtx, audy, sr___, shogun-buildbot, @sonney2k, yoh_, CIA-52, naywhayare, wiking02:58
-!- gsomix [~gsomix@178.45.72.194] has joined #shogun08:23
blackburnokay09:39
@sonney2kblackburn, so09:59
@sonney2kthe 1,1 hack is also suboptimal09:59
blackburnsonney2k: yes because it is needed for each field10:00
@sonney2kblackburn, yeah10:00
@sonney2kand btw the problem exists for sgvector too!10:00
blackburnI do not understand why not to make it (true)10:00
@sonney2kand other referenced data10:00
blackburnby default10:00
blackburnwhat do we lose there?10:00
@sonney2kwe might just get some crashers10:00
blackburnwhy?10:00
@sonney2kbut if we do it we should do it for all SGReferencedData10:00
blackburnwhat should crash?10:01
@sonney2kblackburn, it could be that we relied on this10:01
@sonney2ke.g.10:01
@sonney2kdata pointer is set later10:01
@sonney2kbut that is not intended to be freed later10:01
blackburnit sounds like a hack10:02
blackburndo we really use it somewhere?10:02
@sonney2kno idea10:02
@sonney2kwe can try10:02
blackburnI do not use it anywhre10:02
@sonney2kbut the real patch would just be to remove the 'false'10:02
@sonney2ki.e.10:02
@sonney2kSGMatrix() : SGReferencedData()10:02
@sonney2ketc10:02
blackburnsonney2k: other option is to make serialization take care of ref counts10:02
blackburnit is kind of pain yet10:03
@sonney2kyeah but that we should do later10:03
@sonney2kwhen we do the CMap trick before serializing objects10:04
@sonney2kfor now - we can give it a try10:04
blackburnyes but no time for that now10:04
blackburnargh then we still can release10:04
blackburn:D10:04
@sonney2kmaybe it is more consistent anyways to have refcounting=true always10:04
blackburnyes it sounds more consistent for me10:04
@sonney2klet me just do it...10:05
blackburnfeel free, I can patch shogun only through github web now :)10:05
@sonney2kblackburn, is there an online vim on github :D10:06
@sonney2kblackburn, btw we don't have SGNDArray converted to referenced yet...10:06
blackburnnot the terrible issue10:07
@sonney2kand string / stringlist10:07
CIA-52shogun: Soeren Sonnenburg master * rbb0907c / (6 files in 2 dirs): enable reference counting by default (even in default constructors) for SG Datatypes - http://git.io/NRjBOg10:10
@sonney2kblackburn, gsomix I finally managed to get the infrastructure going for the new website10:10
@sonney2kI put it here for now http://shogun.7nn.de/10:10
@sonney2k(to not kill the old one)10:10
gsomixsonney2k, cool10:14
shogun-buildbotbuild #390 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/390  blamelist: Soeren Sonnenburg <sonne@debian.org>10:16
@sonney2kblackburn, ^ see - some stuff seems to rely on that behavior10:19
@sonney2kanyway I guess we should rather fix the new crashes than doing it the old way...10:19
@sonney2kgsomix, can you help fixing these?10:22
gsomixsonney2k, hm, what's wrong?10:23
blackburnsonney2k: yes, lets rather fix these crashers10:25
blackburngsomix: yeah we'd appreciate if you localize it ;)10:25
gsomixok10:25
CIA-52shogun: Soeren Sonnenburg master * rfc5a120 / src/shogun/features/streaming/StreamingDenseFeatures.cpp : disable refcounting for 'current' vector in streaming features - http://git.io/DWJkNw10:46
blackburnsonney2k: parsebuffer should be fixed still10:47
@sonney2kblackburn, ?10:47
@sonney2kuhh!10:48
@sonney2kCKNN::~CKNN()10:48
@sonney2k{10:48
@sonney2k    SG_FREE(m_train_labels.vector);10:48
@sonney2k}10:48
@sonney2kwtf?!10:48
blackburnhahahah10:49
blackburnyes10:49
blackburnWAT10:49
blackburnlol10:49
@sonney2kblackburn, any idea why we even copy the train labels?10:50
@sonney2kahh we need it to classify later10:50
blackburnsonney2k: to subtract min label10:51
blackburnpretty idiot thing actually10:51
shogun-buildbotbuild #492 of deb3 - modular_interfaces is complete: Failure [failed test libshogun]  Build details are at http://www.shogun-toolbox.org/buildbot/builders/deb3%20-%20modular_interfaces/builds/492  blamelist: Soeren Sonnenburg <sonne@debian.org>10:52
CIA-52shogun: Soeren Sonnenburg master * r2535f44 / src/shogun/multiclass/KNN.cpp : remove free() in destructor (causing a double free) - http://git.io/ybQ7Lg10:56
@sonney2kblackburn, what did you mean 'parsebuffer should be fixed still?'10:57
blackburnsonney2k: it handles vectors in pretty crappy way10:58
@sonney2kblackburn, I really think you should file all the findings we just had as github issues10:59
@sonney2kotherwise this is just forgotten10:59
blackburnsonney2k: I asked gsomix to fix that :)10:59
blackburnwhat else is to do?10:59
@sonney2kblackburn, sparse stuff still fails - but I don't get it. do you have an idea?10:59
@sonney2k==12560== Invalid read of size 811:00
@sonney2k==12560==    at 0xEA6ECB2: shogun::CSparseFeatures<double>::free_sparse_feature_matrix() (SparseFeatures.cpp:95)11:00
@sonney2k==12560==    by 0xEA6F598: shogun::CSparseFeatures<double>::free_sparse_features() (SparseFeatures.cpp:104)11:00
@sonney2k==12560==    by 0xEA6F5E3: shogun::CSparseFeatures<double>::~CSparseFeatures() (SparseFeatures.cpp:90)11:00
@sonney2k==12560==    by 0xEA6F628: shogun::CSparseFeatures<double>::~CSparseFeatures() (SparseFeatures.cpp:91)11:00
@sonney2k==12560==    by 0xE3AF4BC: shogun::CSGObject::unref() (SGObject.cpp:178)11:00
@sonney2k==12560==    by 0xD5982B2: _wrap_delete_SparseRealFeatures (modshogun_wrap.cxx:326578)11:00
@sonney2k==12560==    by 0xD845AA3: _wrap_delete_SparseRealFeatures_closure (modshogun_wrap.cxx:607465)11:00
@sonney2k==12560==    by 0x47C781: frame_dealloc.14762 (in /usr/bin/python2.7)11:00
@sonney2k==12560==    by 0x4B7C81: PyEval_EvalCodeEx (in /usr/bin/python2.7)11:00
@sonney2k==12560==    by 0x4C77DB: function_call.15201.3848 (in /usr/bin/python2.7)11:00
@sonney2k==12560==    by 0x4B1CD0: PyEval_EvalFrameEx (in /usr/bin/python2.7)11:00
@sonney2k==12560==    by 0x4B7C67: PyEval_EvalCodeEx (in /usr/bin/python2.7)11:00
@sonney2k==12560==  Address 0xb0bf3c0 is 0 bytes inside a block of size 96 free'd11:00
@sonney2k==12560==    at 0x4C27D4E: free (vg_replace_malloc.c:427)11:00
@sonney2k==12560==    by 0xD952BE0: shogun::SGSparseMatrix<double>::free_data() (SGSparseMatrix.h:103)11:00
@sonney2k==12560==    by 0xD8D2184: shogun::SGReferencedData::unref() (SGReferencedData.h:157)11:00
@sonney2k==12560==    by 0xD8FDB6A: shogun::SGSparseMatrix<double>::~SGSparseMatrix() (SGSparseMatrix.h:64)11:00
@sonney2k==12560==    by 0xD5992BE: _wrap_SparseRealFeatures_get_sparse_feature_matrix (modshogun_wrap.cxx:326920)11:00
@sonney2k==12560==    by 0x4B01FC: PyEval_EvalFrameEx (in /usr/bin/python2.7)11:01
@sonney2k==12560==    by 0x4B7C67: PyEval_EvalCodeEx (in /usr/bin/python2.7)11:01
@sonney2k==12560==    by 0x4C77DB: function_call.15201.3848 (in /usr/bin/python2.7)11:01
@sonney2k==12560==    by 0x4B1CD0: PyEval_EvalFrameEx (in /usr/bin/python2.7)11:01
@sonney2k==12560==    by 0x4B7C67: PyEval_EvalCodeEx (in /usr/bin/python2.7)11:01
@sonney2k==12560==    by 0x55F2E2: run_mod.42411.3586 (in /usr/bin/python2.7)11:01
@sonney2k==12560==    by 0x4555E4: PyRun_FileExFlags (in /11:01
blackburnhuh11:01
blackburnsonney2k: yes11:03
@sonney2kblackburn, ?11:03
@sonney2kcare to share?11:03
blackburnhttps://github.com/shogun-toolbox/shogun/blob/master/src/shogun/features/SparseFeatures.cpp:37311:03
blackburnwe need to inc ref counter initially11:03
blackburnit should be 2 on creation11:03
@sonney2k    SGSparseMatrix<ST> sm=SGSparseMatrix<ST>(NULL, 0, 0, false);11:06
@sonney2kso that is the fix right?11:06
blackburnactually yes11:06
blackburncould be the fix as well11:06
@sonney2kblackburn, the other would require that we use SGSparseMatrix for SparseFeatures (to be put on todo)11:07
blackburnyes, but refcounter = 2 initially would work as well11:07
blackburnhowever it is a crappy solution :)11:07
CIA-52shogun: Soeren Sonnenburg master * r958ce8f / src/shogun/features/SparseFeatures.cpp : disable refcounting for sparse matrix - http://git.io/tn9YpQ11:07
@sonney2kthat should do it11:08
* sonney2k is hunting some food11:08
blackburnyeah11:08
blackburnI need to hunt for some tea11:08
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has quit [Ping timeout: 248 seconds]11:18
shogun-buildbotbuild #493 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/493  blamelist: Soeren Sonnenburg <sonne@debian.org>11:19
-!- zxtx [~zv@cpe-75-83-151-252.socal.res.rr.com] has joined #shogun11:20
shogun-buildbotbuild #391 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/391  blamelist: Soeren Sonnenburg <sonne@debian.org>11:21
shogun-buildbotbuild #494 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/49411:55
@sonney2kshogun-buildbot, I told you!11:55
shogun-buildbotWhat you say!11:55
shogun-buildbotbuild #392 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/39212:01
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has joined #shogun12:13
-!- satanach [~satanach@137.Red-83-39-123.dynamicIP.rima-tde.net] has joined #shogun12:31
-!- satanach [~satanach@137.Red-83-39-123.dynamicIP.rima-tde.net] has left #shogun []12:34
@sonney2kn4nd0, if you have some time - could you try to get all valgrind errors down to 0 in examples/libshogun ?13:09
CIA-52shogun: Soeren Sonnenburg master * r9e92392 / (3 files in 3 dirs): fix a couple of memory leaks in so example (mostly unnecessary SG_REF's - http://git.io/6kEr9A13:32
-!- Netsplit *.net <-> *.split quits: zxtx13:36
blackburnsonney2k: we have to wait for puffin's patch still13:37
-!- Netsplit over, joins: zxtx13:43
-!- CIA-52 [cia@198.71.88.9] has quit [Ping timeout: 244 seconds]14:17
-!- CIA-30 [cia@cia.vc] has joined #shogun14:27
blackburnsonney2k: we can't have eigen in h, right?14:48
blackburn*.h14:48
-!- Alberto [96a13b06@gateway/web/freenode/ip.150.161.59.6] has joined #shogun14:53
-!- Alberto [96a13b06@gateway/web/freenode/ip.150.161.59.6] has quit [Client Quit]14:53
-!- Alberto [96a13b06@gateway/web/freenode/ip.150.161.59.6] has joined #shogun14:54
-!- CIA-30 [cia@cia.vc] has quit [Ping timeout: 244 seconds]15:08
-!- CIA-27 [cia@cia.vc] has joined #shogun15:09
-!- CIA-27 [cia@cia.vc] has quit [Ping timeout: 244 seconds]15:17
-!- CIA-30 [cia@cia.vc] has joined #shogun15:24
-!- n4nd0 [53b32c87@gateway/web/freenode/ip.83.179.44.135] has quit [Quit: Page closed]15:25
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun15:28
n4nd0naywhayare: hey! how is it going with the KNN thing?15:29
-!- Alberto [96a13b06@gateway/web/freenode/ip.150.161.59.6] has quit [Ping timeout: 245 seconds]15:44
-!- koen-shogun [~ksande@195.114.233.130] has joined #shogun15:45
-!- koen-shogun [~ksande@195.114.233.130] has left #shogun []15:51
n4nd0hi blackburn16:07
blackburnhey16:07
n4nd0have you heard about something called categorical prediction?16:07
n4nd0I'm reading a book about large margin methods for ASR and they present this16:08
n4nd0at first, I thought that it was going to be structured prediction16:08
n4nd0but I am reading the chapter were categorical prediction is introduced and I don't understand a sh**16:08
n4nd0:D16:08
n4nd0anybody there?16:27
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]16:46
-!- hoijui [~hoijui@141.23.65.251] has joined #shogun16:58
-!- hoijui [~hoijui@141.23.65.251] has quit [Quit: Leaving]17:13
@sonney2kblackburn, yes - no eigen in .h18:37
@sonney2kblackburn, but usually that is not a problem (you can move any function code to -> .cpp)18:37
@sonney2kblackburn, btw same thing holds for pthreads etc18:37
@sonney2kblackburn, can you fix the memory leaks in the libshogun examples? there are quite many...18:38
@sonney2kwiking, any thoughts on getting proper gtest/gmock detection?18:39
-!- blackburn1 [~blackburn@188.168.5.103] has joined #shogun19:11
@sonney2kblackburn1 seen this:19:16
@sonney2k<sonney2k> blackburn, yes - no eigen in .h19:17
@sonney2k<sonney2k> blackburn, but usually that is not a problem (you can move any function code to -> .cpp)19:17
@sonney2k<sonney2k> blackburn, btw same thing holds for pthreads etc19:17
@sonney2k<sonney2k> blackburn, can you fix the memory leaks in the libshogun examples? there are quite many...19:17
blackburn1sonney2k: yeah thanks19:17
blackburn1sonney2k: about eigen - it is the issue with puffin's code  so it blocks us to release just now19:19
blackburn1he is working on  that19:20
blackburn1sonney2k: I will a little later19:25
blackburn1(fix)19:25
-!- CIA-30 [cia@cia.vc] has quit [Ping timeout: 244 seconds]19:39
-!- blackburn1 [~blackburn@188.168.5.103] has quit [Ping timeout: 252 seconds]19:43
-!- CIA-30 [cia@cia.vc] has joined #shogun20:09
-!- CIA-30 [cia@cia.vc] has quit [Ping timeout: 252 seconds]20:39
-!- CIA-33 [cia@cia.vc] has joined #shogun21:15
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has joined #shogun21:20
-!- blackburn666 [d55786d5@gateway/web/freenode/ip.213.87.134.213] has joined #shogun21:21
n4nd0blackburn666: 666? feeling diabolic? :P21:23
blackburn666n4nd0: exactly21:23
n4nd0blackburn666: let me ask you21:23
n4nd0I think something has gotten broken kind of badly in my computer21:24
n4nd0whenever I want to execute an example of libshogun I get21:24
n4nd0Segmentation fault (core dumped)21:24
n4nd0have you seen that before?21:24
blackburn666hmm not really21:25
blackburn666just any example fails?21:26
n4nd0I think that basically all21:26
n4nd0I just tried with basic_minimal21:26
n4nd0I am gooing to configure and compile from zero21:26
blackburn666I'd suggest to reinstall from the blank21:27
n4nd0do you mean ./configure, make clean, make and make install?21:28
n4nd0or something else?21:28
blackburn666yeah or even git clean -dfx if you have nothing unstaged for commit21:28
-!- CIA-33 [cia@cia.vc] has quit [Ping timeout: 244 seconds]21:30
-!- n4nd0 [~nando@s83-179-44-135.cust.tele2.se] has quit [Quit: leaving]21:36
-!- CIA-27 [cia@cia.vc] has joined #shogun21:40
@sonney2kblackburn, if you see n4nd0 tell him I need bern4rds help now...22:12
@sonney2kblackburn666, ^22:15
-!- gsomix is now known as blackburnsqrt-122:29
-!- blackburnsqrt-1 is now known as blackburnsqrt[-122:29
blackburnsqrt[-1aww :(22:29
-!- blackburnsqrt[-1 is now known as blackburnsqrt-122:30
-!- blackburnsqrt-1 is now known as gsomix22:30
-!- blackburn666 [d55786d5@gateway/web/freenode/ip.213.87.134.213] has quit [K-Lined]22:47
-!- blackburn [5bdfb203@gateway/web/freenode/ip.91.223.178.3] has quit [K-Lined]22:47
--- Log closed Sat Sep 01 00:00:17 2012