IRC logs of #shogun for Tuesday, 2018-02-13

« Monday, 2018-02-12 Index Wednesday, 2018-02-14 »
--- Log opened Tue Feb 13 00:00:00 2018
-!- Prashanth [uid170915@gateway/web/irccloud.com/x-bxyvvrqhkqezgvme] has joined #shogun04:47
-shogun-buildbot:#shogun- Build nightly_all #94 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/22/builds/9408:06
-!- Prashanth [uid170915@gateway/web/irccloud.com/x-bxyvvrqhkqezgvme] has quit [Quit: Connection closed for inactivity]08:16
-!- HeikoS [~heiko@bmi-airport-1.inf.ethz.ch] has joined #shogun10:28
-!- mode/#shogun [+o HeikoS] by ChanServ10:28
@sukey[https://github.com/shogun-toolbox/shogun] Wiki page: Google-Summer-of-Code-2018-Projects edited by karlnapf10:55
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/1807e418b248b38c34675a6b0e8d451e9e4f7b49 by vigsterkr11:11
-!- travis-ci [~travis-ci@ec2-54-80-214-111.compute-1.amazonaws.com] has joined #shogun11:30
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34087854011:30
-!- travis-ci [~travis-ci@ec2-54-80-214-111.compute-1.amazonaws.com] has left #shogun []11:30
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4157 closed by bmurauer11:41
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4157 reopened by vigsterkr11:42
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4173 opened by bmurauer11:47
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4141 closed by vigsterkr11:47
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/417311:47
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/417311:55
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has joined #shogun12:16
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34087854012:16
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has left #shogun []12:16
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has joined #shogun12:28
travis-ciit's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34052150312:28
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has left #shogun []12:28
Trixiswiking: i've stalled a bit with the java fix d.2. uni work13:18
@wikingdont worry13:18
Trixiswiking: any clue when the next version of shogun will come up on anaconda btw?13:18
@wikingwe had a lot of stuff as well here in the meanwhile :D13:18
Trixisheh fair13:18
@wikingmmm13:18
@wikingwe are planning a new release soonish13:18
@wikingalthough there are many moving parts atm13:19
@wikingso yeah i mean we wont release till end of next week for sure13:19
Trixisahh cool, mostly wanted to know so that i can plan ahead a bit13:20
@wikinglisitsyn, around by any chance?13:22
lisitsynwiking: hey13:22
@wikinggreat13:23
@wikingso i have a question because you touched these stuff in the past13:23
lisitsynaham13:23
lisitsynbut I am markov memory-wise lol13:23
@wikinghttps://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/multiclass/MulticlassLibSVM.cpp#L162-L18813:24
@wiking:)13:24
lisitsynah ok13:24
@wikingdo you know wtf is actually happening here :D13:24
lisitsynwiking: I vaguely remember fixing something like that13:25
@wikingand namely when you get the following labelling: [1,1,2,2,3,3] this one fails here https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/multiclass/MulticlassLibSVM.cpp#L20013:25
lisitsynit is OvO13:25
@wikingyeah i realised that it is the OvO13:25
@wikingi mean why is it here ? :D13:25
lisitsynit looks ugly to me as well13:25
@wikingconceptually13:25
@wikingbut yeah lets not go into this13:25
@wikingbasically if |num class| = 3 shit happens13:25
@wikingdunno what happens actually when |num class| >313:26
@wikingwhen it's. < 3 it works :D13:26
lisitsynwiking: I can understand what's different when num_classes is 213:26
lisitsynbut why 3 is magical I don't really get13:26
@wiking:)13:27
lisitsynwiking: can you try dropping num_classes == 3?13:28
@wiking:D13:28
@wikingi'm just checking how OvO sets m_num_classes13:29
@wikingwhich doesn't seem to be too obvious :D13:29
lisitsynwiking: 2 is special in case of OvO because you don't need it twice13:30
lisitsynis 3 special?13:31
@wikingdunno why13:32
lisitsynwiking: well I can get why 2 is special :)13:32
@wikingbtw bad news is13:33
@wikinghowever, if I change the y array to [1,1,2,2,1,1], no such error occurs. If i put four different classes in y (e.g. [1,1,2,2,3,4]), the error occurs as well.13:33
@wikingbased on the error report :D13:33
@wikinglisitsyn, the most hipster stack: https://attardi.org/pytorch-and-coreml13:33
@wiking;)(13:33
@wikinglisitsyn, hahah train went fine but then we get an assertation error here13:35
@wikinghttps://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/multiclass/MulticlassSVM.cpp#L9913:35
@wikingoh wait13:35
* wiking is an idiot13:35
lisitsynwiking: ?13:36
@wikingjust a sec13:36
@wikingin the meanwhile read about the hipster stack13:36
@wikingreact+coreml+pytorch13:37
@wiking:D13:37
@wikingbtw if i dont consider 3 as a special case13:37
@wikingwe still have the same assertation error13:37
@wikingMulticlassLibSVM.cpp line 200: SVM set failed13:38
@wiking:<?13:38
@wikingok lemme see what is happening in that if thingy13:38
@wiking:D13:38
@wikingas that seems to be some magical land13:38
@wikingoooh13:40
@wikinghttps://pastebin.com/DuiRwkfn13:40
@wikingidx seems to be +113:40
@wikinglol13:42
@wikingnice one13:42
@wikinghttps://github.com/shogun-toolbox/shogun/commit/e13d0fdb731ece63d5a98b6e26682991509399b413:42
@wikingcheck the log commit log13:43
@wiking :)13:43
@wikingand then lisitsyn  you :)13:43
@wikinghttps://github.com/shogun-toolbox/shogun/commit/a55b25e0b87ae2bbc5383a232f26d63ce2fad2ff13:43
@wiking:>13:43
lisitsynoh ok13:43
lisitsynwiking: so you say it never worked? :)13:44
@wikingdunno :D13:44
@wikingi'm just trying to figure out what the hell is happening here13:44
@wiking:>13:44
lisitsynme too13:44
lisitsynit is 6 years13:44
lisitsynago13:44
@wikingnow that finally somebody started to use it :P13:44
@wikingwe should maybe make it work :D13:44
@wikingthis definitely calls for a cookbook later :)13:47
@wikinghttp://shogun.ml/examples/latest/examples/multiclass_classifier/svm.html13:47
@wiking:S13:47
@wikingi guess this one uses 2 class ? :D13:47
@wikingmmm no13:49
@wikingit has 4 labels13:49
@wiking:(13:49
@wikingooooh13:49
@wikinglemme just try something13:49
@wikingok13:52
@wikinglisitsyn, you know what is the problemo :D13:52
@wikingy = np.array([1, 1, 2, 2, 3, 3], np.float64)13:52
@wikingif i change it to y = np.array([0, 0, 1, 1, 2, 2], np.float64)13:52
@wikingi.e. it's 0 indexed13:52
@wikingthen it's all good13:53
lisitsynit works then?13:53
@wiking:D13:53
lisitsynok I see13:53
lisitsynhack based on 013:53
@wikingso i mean what would happen13:53
lisitsynwiking: I mean I wasn't like that these days13:53
lisitsynI surely checked if it worked13:53
lisitsyn:D13:53
lisitsynbut with 013:53
@wikingyeye ynono13:53
lisitsynwiking: actually13:53
@wikingyou see13:53
lisitsynI think13:53
@wikingwe have a test13:53
@wikingfor this13:53
@wiking:)13:53
@wikingbut it is all 0 based indexing13:53
lisitsynwe rely on 0 based13:53
lisitsynquite alot13:53
@wikingit doesn't allow you arbitrary13:53
lisitsynI think a lot of things crash13:53
@wikinglabeling13:53
lisitsynif you don't use 0-based13:54
@wikingwhich is of course stupid13:54
@wikingbecause in case of multiclass it'd be good to have13:54
@wikingeven13:54
lisitsynindeed but we use arrays instead of maps13:54
lisitsynthat's why we rely on 013:54
@wiking["egg", "vodka", "mushroom"]13:54
@wikinglabeling13:54
@wiking:)13:54
@wikingi mean we should just hide this internally i guess :)13:55
lisitsynwiking: true13:55
@wikingsome places13:55
@wikingthis is done13:55
@wikinglike in RF13:55
@wikingor bagging machine per se13:55
@wikingi know that there was a confusion there13:55
@wikingabout the index and value13:55
@wikingand i've fixed that 6 months ago13:55
@wikingbut i guess we have a contamination of concepts here13:55
@wikingwith labels used as indicies13:55
@wikingTrixis, this can be closed or https://github.com/shogun-toolbox/shogun/issues/412014:19
@wiking?14:19
-!- ironstark [uid221607@gateway/web/irccloud.com/x-qmdiqxzgjjyivdos] has quit []14:47
-!- ironstark [sid221607@gateway/web/irccloud.com/x-fsyiumzzdbyrtjlm] has joined #shogun14:48
@sukey[https://github.com/shogun-toolbox/shogun] vigsterkr pushed 3 commits:14:57
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/88993f396e4668f912a2e45f3fd9dde6a089e4ec14:57
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/a4dfb372271c56036c1bccf3004ca3b197f82bdb14:57
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/431e56b67344e0662444c0efbd4aefde47990fa414:57
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has joined #shogun15:01
travis-ciit's Sergey Lisitsyn's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34052150315:01
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has left #shogun []15:01
@sukey[https://github.com/shogun-toolbox/shogun] vigsterkr pushed 3 commits:15:13
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/88993f396e4668f912a2e45f3fd9dde6a089e4ec15:13
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/a4dfb372271c56036c1bccf3004ca3b197f82bdb15:13
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/431e56b67344e0662444c0efbd4aefde47990fa415:13
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4157 closed by vigsterkr15:14
Trixiswiking: iirc its still an issue, lemme check15:46
Trixismight have to compile develop for that15:46
Trixiswiking: yep its still an issue15:51
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has joined #shogun16:04
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34096397016:04
-!- travis-ci [~travis-ci@ec2-50-16-107-62.compute-1.amazonaws.com] has left #shogun []16:04
@wikingTrixis, ok then lets keep it open :)16:23
Trixiswiking: i tried to pinpoint whatever function was causing that issue with gdb but it wasnt very informative16:24
-!- travis-ci [~travis-ci@ec2-54-242-129-51.compute-1.amazonaws.com] has joined #shogun16:59
travis-ciit's Viktor Gal's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34096397016:59
-!- travis-ci [~travis-ci@ec2-54-242-129-51.compute-1.amazonaws.com] has left #shogun []16:59
@sukey[https://github.com/shogun-toolbox/shogun] karlnapf pushed 3 commits:17:29
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/3ba681cac7cacf9c90a2928b52d8c5ec8bca7b9017:30
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/7ca11af5011a8b8bcb8085c26bef40e038eb45cb17:30
@sukeyhttps://github.com/shogun-toolbox/shogun/commit/02dfae62753c603f12fce5b9b639e2bad69b681017:30
@wikinglisitsyn, this is a great command cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON ..17:42
@wiking:)17:42
-!- Prashanth [uid170915@gateway/web/irccloud.com/x-zijzehlbalfdptmk] has joined #shogun18:12
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/fefea75743e945845a67a7e8f0dc6e158d29a4f8 by karlnapf18:26
-!- HeikoS [~heiko@bmi-airport-1.inf.ethz.ch] has quit [Quit: Leaving.]18:26
-!- travis-ci [~travis-ci@ec2-54-242-129-51.compute-1.amazonaws.com] has joined #shogun18:52
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34102492818:52
-!- travis-ci [~travis-ci@ec2-54-242-129-51.compute-1.amazonaws.com] has left #shogun []18:52
-!- travis-ci [~travis-ci@ec2-54-242-129-51.compute-1.amazonaws.com] has joined #shogun19:18
travis-ciit's Heiko Strathmann's turn to pay the next round of drinks for the massacre he caused in shogun-toolbox/shogun: https://travis-ci.org/shogun-toolbox/shogun/builds/34104979019:18
-!- travis-ci [~travis-ci@ec2-54-242-129-51.compute-1.amazonaws.com] has left #shogun []19:18
-!- dhazz [7a0f4d8b@gateway/web/freenode/ip.122.15.77.139] has joined #shogun19:23
-!- dhazz [7a0f4d8b@gateway/web/freenode/ip.122.15.77.139] has quit [Ping timeout: 260 seconds]20:13
-!- durovo [~durovo@7d.78.2ea9.ip4.static.sl-reverse.com] has quit [Remote host closed the connection]21:10
-!- durovo [~durovo@cc.b7.2ea9.ip4.static.sl-reverse.com] has joined #shogun21:18
-!- Prashanth [uid170915@gateway/web/irccloud.com/x-zijzehlbalfdptmk] has quit [Quit: Connection closed for inactivity]21:36
@wikinglisitsyn, yo21:42
--- Log closed Wed Feb 14 00:00:01 2018
« Monday, 2018-02-12 Index Wednesday, 2018-02-14 »

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!