IRC logs of #shogun for Wednesday, 2018-01-03

« Tuesday, 2018-01-02 Index Thursday, 2018-01-04 »
--- Log opened Wed Jan 03 00:00:03 2018
-!- kiranmuthigi1 [75d35a9b@gateway/web/freenode/ip.117.211.90.155] has joined #shogun08:22
kiranmuthigi1I installed shogun on my Anaconda and that happened successfully. But when I tried installing in my macOS, I installed Homebrew and then "brew install homebrew/science/shogun". It gave me a  warning "homebrew/science was deprecated. This tap is now empty as all its formulae were migrated." and then an error "No available formula with the name "homebrew/science/shogun" "08:26
kiranmuthigi1Thus, I was not able to install shogun on my macOS, but installed successfully on Anaconda. Please help08:26
-!- kiranmuthigi1 [75d35a9b@gateway/web/freenode/ip.117.211.90.155] has quit [Ping timeout: 260 seconds]08:38
-!- braceletboy [uid256084@gateway/web/irccloud.com/x-xnapwexnihdjywyd] has joined #shogun08:40
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4064 merged by lisitsyn08:57
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/3ac662cca1899a1c5dd6a5ea820854f19b2e8da1 by lisitsyn08:57
@wikinglisitsyn, here?10:05
lisitsynwiking: hey10:06
lisitsynque paso10:07
@wikingso10:08
@wikingi'm trying to get the cereal thing in shape10:08
lisitsynah10:08
@wikingah fuck10:08
@wikingmy b character is stuck in the macbook10:08
@wiking:)10:08
@wikingthis new mac is rather shitty quality10:08
lisitsynlatest macbook are funny10:08
@wikinganyhow so i was looking at this and i feel a bit funny10:08
@wikinghttps://github.com/shogun-toolbox/shogun/blob/feature/cereal/src/shogun/lib/any.h#L5010:09
lisitsynyikes10:09
lisitsynit looks like the old serialization framework10:10
lisitsyn:D10:10
@wikingyeah10:11
@wikingso i was trying to figure out something to use10:12
@wikingtypeid10:12
@wikingbut yeah that's no goodie :(10:12
@wikingi mean what i can do is i guess10:12
@wikingsomething similar10:12
@wikinglike your patch10:12
@wiking:)10:12
lisitsynwell10:12
@wikingfor sgmatrix and sgvector10:12
lisitsynI think one principle to take10:12
lisitsynis to do everything inside policy10:12
lisitsynthis way you already know the type10:12
lisitsynthe policy knows the type10:13
@wikingmmm does it? :)10:14
lisitsynyes10:14
lisitsynthat's the trick10:14
lisitsynpolicy is <T> but base class is non-generic10:15
lisitsynso if you can pass it through the non-generic interface to <T> policy10:15
lisitsynyou win10:15
lisitsyn:D10:15
lisitsynwiking: see what I mean?10:18
@wikingjust a sec10:18
@wikingin a call10:18
lisitsynokie10:18
-!- travis-ci [~travis-ci@ec2-54-163-106-182.compute-1.amazonaws.com] has joined #shogun10:20
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/32448293010:20
-!- travis-ci [~travis-ci@ec2-54-163-106-182.compute-1.amazonaws.com] has left #shogun []10:20
lisitsynbazdmeg10:20
lisitsynyou coward travis-ci10:21
lisitsynok so what I mean10:24
lisitsynyou call compare from policy10:24
lisitsynso if you want to try to do the same with ar10:24
lisitsynyou gotta call it from inside the policy as well10:25
@wiking:D10:25
@wikinglisitsyn, i'm just trying to figure out how i can get a factory method10:26
lisitsynfactory?10:26
@wikingwhen deserializing10:26
lisitsynah10:27
@wikingso i mean when i serialize stuff10:27
@wikingnot only in need the type10:27
@wikingbut then i need a factory to generate the right type10:27
@wikingor?10:27
@wiking:)10:27
lisitsynyeah kind of10:28
@wikingi would need soemthing like10:29
@wikingcreate_object<SGMatrix>(type);10:29
@wiking:)10:29
@wikingbut not really10:29
@wikingbecause sgmatrix is typed right...10:29
lisitsynhow does cereal do that?10:30
@wikingwhat? :)10:32
@wikingah10:32
@wikingthere's no support yet for std::any in cereal10:32
lisitsynno, how does it call the proper method to create an instance10:33
@wikinglemme check10:34
@wikingno idea10:34
lisitsynah it does not10:35
lisitsynwiking: that's what we do in cereal_load10:35
@wikingi mean yeah if the type is not defined10:35
@wikingi mean supported by cereal10:35
@wikingyou have to specify your own serialization10:36
@wikingmethod10:36
lisitsynmeh10:36
lisitsynthat's not pretty this way now10:36
@wikingso we need to define here the cereal_load/save10:36
lisitsynwe gonna touch the any everytime we need something10:36
@wiking:)10:36
@wikingunit test will break10:36
@wikingso it's ok :)10:36
@wikingwe just need to add versioninng10:37
@wikingto the serialization format10:37
@wikingand that's all :)10:37
lisitsynno I mean it is recompile-everything10:37
@wikingyeah well10:37
@wikinglife is shit :)10:37
lisitsynI think they should be separate10:37
@wikingthe serialization? :)10:37
lisitsynyes10:38
lisitsynany <-> serialization10:38
lisitsynnot anyserializationmodelselection10:38
lisitsyn:D10:38
lisitsynwiking: ah btw a ltitle off-topic10:39
lisitsyndid you know shogun CRASHED in case of exception10:39
@wikingyes yes10:40
@wikingi know that exception handleing10:40
lisitsynI fixed that10:40
@wikingwas crashing10:40
lisitsynin python10:40
@wikingthe python interpreter10:40
@wikingyeap10:40
@wikingi've seen it ;)10:40
lisitsynbut we need the same thing everywhere10:40
@wikingthat is nice fix there10:40
lisitsynI was like ashamed realizing that10:40
lisitsynI mean CRASHING KABOOM10:40
lisitsynhow come10:40
lisitsyn:D10:40
@wikingyeah10:40
@wikingit was always like this10:41
@wiking:))))10:41
lisitsynwell we discussed wrong things then10:41
lisitsyn:D10:41
@wiking:P10:41
@wikingyes10:41
@wikingindeed :)10:41
@wikinganyhow10:41
lisitsynok back to cereal10:41
lisitsynuhm10:41
@wikinglemme just try to imagine10:42
@wikinghow to decouple cerelization ;P10:42
lisitsynyou know I WAS a fan of template stuff10:42
lisitsyn:D10:42
@wiking:E10:42
@wiking:D10:42
@wikingi mean the whole problem already with cereal10:44
@wikingis that we tend to hide stuff in implementation10:44
@wikingwhereas this kills us in a way because cereal meant to be able to handle any Archive format (see template arg)10:44
@wikingpersonally i dont like this10:45
@wikingbut then if we start doing it real10:45
@wikingthen we need to include cereal everywhere10:45
@wikingin base class headers10:45
@wiking:(10:45
@wikinganyhow still i dont see how do you wanna decouple any's serialization10:45
lisitsynI don't know how10:45
lisitsynbut we need to10:46
lisitsyn:P10:46
@wikingbecause afaik cereal needs the serialization for a given class to be defined/implemented in the class10:46
lisitsynwiking: uhm mmm10:48
lisitsynhttps://www.youtube.com/watch?v=iNLsn_LfcV410:49
@wiking:D10:49
lisitsynuhmm10:50
lisitsyn:D10:50
lisitsynwiking: so Archive is templated because10:50
lisitsynArchive might be json10:50
lisitsynor something else10:51
lisitsyn(honestly I don't think it is a great idea)10:51
lisitsynpero es tarde10:51
@wikingi mean we can actually fix it to 3 types10:53
@wikingjson, xml, binary10:53
lisitsynwiking: do you think working with typeid might be too slow for us?10:54
@wikingin what sense?10:54
lisitsynyou can easily obtain typeid from any10:54
lisitsynit has the method to do that10:54
@wikingyeah but what do i do with typeid? :)10:54
@wikingif i use that10:54
lisitsynswitch it10:54
@wikingi still need to somewhere implement10:54
@wikinga factory10:55
@wiking:)10:55
@wikingcan you actually10:55
lisitsynyes but this would be separate from any10:55
@wikingdecltype(typeid) ?10:55
lisitsynno :D10:55
@wikingi guess no :)10:55
@wikingok so what i could do is create a factory for SGVector and SGMatrix10:55
@wikingfor the rest we have a factory10:55
@wiking:)10:55
@wikingshogun::create10:56
@wikingi mean of course10:56
@wikingmmm10:56
@wikingany can never hold SGObject10:56
@wikingor?10:56
lisitsynit can hold SGObject*10:56
@wikingmmm10:56
@wikingok then we are a bit fucked :)10:56
lisitsynyes10:56
@wikingi mean then if we use typeid10:57
@wikingthen we dont have a factory method10:57
@wiking:P10:57
@wikingwhereas if we use any.name()10:57
@wikingwe do10:57
@wikingright?10:57
lisitsynwhat's name?10:57
@wikingmmm10:57
@wikingno10:57
@wikingtype()10:57
@wikingbut it's not good10:57
lisitsynah10:57
@wikingbecause it's using10:57
@wikingdemangle10:57
@wikingwhat i'm thiking atm10:57
lisitsynyou can have it mangled10:57
lisitsyn:)10:57
@wikingthat maybe the best is10:58
@wikingto extend10:58
@wikingshogun::create10:58
@wikingwith SGVector and SGMatrix10:58
@wiking:PP10:58
lisitsynah yeah why not10:58
@wikingand then in any cerializtion10:58
@wiking*serialization10:58
-!- kiranmuthigi1 [2a684d92@gateway/web/freenode/ip.42.104.77.146] has joined #shogun10:58
@wikingwe use get_name10:58
@wikingin case it's SGObject10:59
lisitsynwiking: do you think we can specialize for these 3 types of Archive?10:59
@wikinglisitsyn, atm it's like this10:59
@wikingwe only support json,xml,binary10:59
@wikingwith cereal10:59
lisitsynok then I know how do that10:59
@wikingmmmmm11:00
@wikinglisitsyn, !!!!11:00
lisitsynwiking: we just add ar(Json), ar(Binary), ar(Xml) to the policy base class11:00
lisitsynand implement it in policies11:00
lisitsynnot a big deal :P11:00
@wikinghttps://github.com/shogun-toolbox/shogun/blob/feature/meta_example_r_fix/src/shogun/io/Serializable.h11:00
@wikingwe could actually use this11:01
@wiking:P11:01
@wikingbecause this wraps11:01
@wikingSGMatrix, SGVector11:01
@wikingand primitive types11:01
@wikingto SGObject11:01
@wiking:PPP11:01
lisitsynyeah but how's that helpful? :)11:01
lisitsynah11:01
lisitsynI see11:01
lisitsynyeah kind of11:01
@wikingso i mean we just use this11:01
@wikingwhen we serialize11:01
@wikingand the same story to deserialize11:02
@wikingno?11:02
lisitsynnot that easy because you'd have to switch on deser11:02
lisitsynwiking: ok I know how to split them11:03
lisitsyn:D11:03
@wikinglisitsyn, describe or do it directly11:03
@wikingin the feature branch11:04
lisitsynnah I don't do feature branches you know11:04
lisitsyn:D11:04
lisitsynlet me gist11:04
@wikingi'm thikning of getting that Serializable class out of that feature branch11:04
@wikingto get rid of wrapped11:04
lisitsynwiking: https://gist.github.com/lisitsyn/652cb77dd86c87d81545fb82e9d3d07011:06
lisitsynaha?11:06
lisitsynthat's much better I believe11:07
@wikingmmm11:07
@wikinglemme see how that works with11:07
@wikingcerela11:07
@wikingbut i get the idea11:07
@wikingbut first11:07
@wikinglemme factor out the Serializable11:07
lisitsynyou implement it in class Cereal11:07
@wikingwanna get rid of11:07
@wikingWrapped*11:07
@wikingas that was problem11:08
@wikingmany times already11:08
@wiking;P11:08
@wikingbut yeah i see what you mean'11:08
@wikingbut yeah as soon as we have11:08
@wikingcereal11:08
@wikingand we can drop the whole SerializeableFile*11:08
@wikingwe can start dropping things like11:08
@wikingdynamic_object_array11:08
@wiking:)11:08
@wikingas cereal has support for STLs11:08
lisitsynI can add serialize thingy11:08
lisitsynto any11:09
lisitsynI mean just an interface11:09
@wikingcool11:09
lisitsynand its call11:09
@wikingdo it11:09
@wikingi can fill eout the rest11:09
lisitsynok but a bit later today11:09
lisitsynyou just inherit from that thing later11:09
lisitsynand then it works11:09
lisitsyn:P11:09
lisitsynbut not deserialization11:09
lisitsynthough it is doable as well11:09
@wiking:>11:09
lisitsynyou just read the value and wrap it to any11:09
lisitsynwiking: I am thinking of renaming any to Value11:30
lisitsyn:D11:30
lisitsynI mean it becomes quite a powerful class11:30
lisitsynit is like any but more complex11:30
-!- kiranmuthigi1 [2a684d92@gateway/web/freenode/ip.42.104.77.146] has quit [Ping timeout: 260 seconds]11:35
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4065 opened by lisitsyn12:28
lisitsynwiking: ^ here you go12:28
@wiking\p/13:19
@wikinggoing to lunch13:19
@wikingbbl13:19
Trixisbtw is there a known issue with shogun on python 3.5.4? Created conda env with python3.5 and shogun, segfaults with code 11 whenever import shogun is called.14:04
@wikingTrixis, which versiomn?14:18
Trixisshogun                    6.1.3           py35_blas_openblas_0  [blas_openblas]  conda-forge14:19
Trixisshogun-cpp                6.1.3           blas_openblas_0  [blas_openblas]  conda-forge14:19
@wikingmmmmm14:30
@wikingstrangeo14:30
Trixis"conda create -n python35 python=3.5 anaconda" -> "source activate python35" -> "conda install -c conda-forge shogun" was used to create the env14:33
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4065 merged by lisitsyn14:57
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/b32e658ed0a5f0d9c32fee5b873a52dedb67518d by lisitsyn14:57
Trixisi would assume it's because either shogun or some of its deps are compiled for different python version14:59
@wikingTrixis, need to check15:42
@wikingwhat i'm using15:42
Trixiskk15:42
Trixisthanks15:42
@wikingsimple import shogun fails?15:43
Trixisyep15:44
Trixisjust immediately returns with segfault cold 1115:45
Trixis*code 1115:45
Trixistried importing numpy, etc to make sure it wasnt a larger issue, that all works15:45
@wikinglisitsyn, what about the other types? :)15:48
@wikingint8? :P15:48
@wikingTrixis,15:51
@wiking(shogun-test35) [galv@lm-login-01 ~]$ python15:51
@wikingPython 3.5.4 |Anaconda custom (64-bit)| (default, Oct 13 2017, 11:22:58)15:51
@wiking[GCC 7.2.0] on linux15:51
@wikingType "help", "copyright", "credits" or "license" for more information.15:51
@wiking>>> import shogun15:51
@wiking>>>15:51
@wikingworksforme ?15:51
TrixisPython 3.5.4 |Anaconda, Inc.| (default, Nov  8 2017, 18:11:28)15:52
Trixis[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)] on darwin15:52
TrixisType "help", "copyright", "credits" or "license" for more information.15:52
Trixis>>> import shogun15:52
TrixisSegmentation fault: 1115:52
lisitsynSegmentation fault!!111115:53
lisitsynsorry :P15:53
lisitsynwiking: add more types if needed :P15:53
@wikingTrixis, lldb/gdb ?15:57
Trixislldb i  think15:57
Trixisremind me how do i check?15:57
@wikinglldb python16:00
-!- travis-ci [~travis-ci@ec2-54-242-116-186.compute-1.amazonaws.com] has joined #shogun16:12
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/32458767516:12
-!- travis-ci [~travis-ci@ec2-54-242-116-186.compute-1.amazonaws.com] has left #shogun []16:12
-!- travis-ci [~travis-ci@ec2-54-163-106-182.compute-1.amazonaws.com] has joined #shogun16:37
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/32458767516:37
-!- travis-ci [~travis-ci@ec2-54-163-106-182.compute-1.amazonaws.com] has left #shogun []16:37
Trixiswiking: https://pastebin.com/pqhQeF6916:59
@wikingchecking16:59
-!- kiranmuthigi1 [2a684d92@gateway/web/freenode/ip.42.104.77.146] has joined #shogun17:04
kiranmuthigi1lisitsyn... I installed shogun on my Anaconda and that happened successfully. But when I tried installing in my macOS, I installed Homebrew and then "brew install homebrew/science/shogun". It gave me a  warning "homebrew/science was deprecated. This tap is now empty as all its formulae were migrated." and then an error "No available formula with the name "homebrew/science/shogun" "  Thus, I was not able to install shogun on my 17:06
@wikingTrixis, bt?17:06
Trixishm?17:08
@wikingtype bt in lldb17:09
@wikingand paste it into pastebin17:09
@wikingi.e. backtraxe17:09
@wiking*backtrace17:09
@wikingkiranmuthigi1, use conda17:10
@wikingkiranmuthigi1, we are still working on the brew migration from science17:10
@wikingsee https://github.com/Homebrew/homebrew-core/pull/2144417:11
kiranmuthigi1wiking: ok Thanks:)17:11
Trixishttps://pastebin.com/589nEdHq wiking17:16
@wikingmmm17:21
@wikinginteresting17:21
@wikingTrixis, and have u tried with py27 or py36?17:25
Trixiswiking: any ideas17:26
@wikingnot really17:29
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/34490ad4fa30543bd9e2b316ac02f3ca4a58c288 by lisitsyn17:32
Trixiswiking: guess i could get conda to install older version of shogun...17:35
@wikingmmm i wonder if that changes anything17:35
Trixiswiking: yep that fixes it17:37
@wiking:O17:38
@wikingwhich version?17:38
Trixisconda install -c conda-forge shogun=6.0.017:39
Trixisso id assume 6.0.017:39
@wikingmmm17:40
@wikingcan you maybe try other versions?17:40
@wikingthere was 6.1.017:40
@wiking6.1.217:40
@wikingat least17:40
@wiking:)17:40
Trixishttps://anaconda.org/conda-forge/shogun/files i only saw 6.0.0 here17:40
Trixissec gonna try 6.1.217:40
-!- kiranmuthigi1 [2a684d92@gateway/web/freenode/ip.42.104.77.146] has quit [Ping timeout: 260 seconds]17:40
Trixisyeah that one wasnt found17:40
@wiking6.1.117:41
TrixisUnsatisfiableError: The following specifications were found to be in conflict:17:43
Trixis  - python 3.5*17:43
Trixis  - shogun 6.1.1* -> python 3.6*17:43
@wiking?17:43
@wikingos?17:43
Trixisand unfortunately python 3.6 is incompatible with other libraries17:43
Trixisthats for shogun 6.1.117:43
@wikingwhich os?17:43
Trixisright now OSX, later on linux17:44
@wikinghttps://anaconda.org/conda-forge/shogun/6.1.1/download/linux-64/shogun-6.1.1-py35_blas_openblas_0.tar.bz217:44
@wikingbut yeah there's no osx17:44
braceletboyHey!!  @wiking and @trixis, I am working on an issue: https://github.com/shogun-toolbox/shogun/issues/4049 I had some doubts and posted my doubts in that page, but didn't receive any replies. Can anyone help me here?17:44
@wikingof 6.1.117:44
@wikingTrixis, can you try 6.1.3 on linux?17:44
Trixiscant test linux yet because i think the machines are under maintenance rn17:44
Trixislemme check17:44
Trixisright yeah cant ssh in17:48
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/dd97dba2b10fa3e9407ac1cea7a795d64e26f927 by lisitsyn17:53
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/ee3aa74f81c01afa65782f478ce135e1ec8023d0 by lisitsyn17:59
-!- travis-ci [~travis-ci@ec2-54-163-106-182.compute-1.amazonaws.com] has joined #shogun18:49
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/32465165518:49
-!- travis-ci [~travis-ci@ec2-54-163-106-182.compute-1.amazonaws.com] has left #shogun []18:49
-!- travis-ci [~travis-ci@ec2-54-242-116-186.compute-1.amazonaws.com] has joined #shogun19:13
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/32465165519:13
-!- travis-ci [~travis-ci@ec2-54-242-116-186.compute-1.amazonaws.com] has left #shogun []19:13
-!- travis-ci [~travis-ci@ec2-54-242-116-186.compute-1.amazonaws.com] has joined #shogun19:45
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/32465946919:45
-!- travis-ci [~travis-ci@ec2-54-242-116-186.compute-1.amazonaws.com] has left #shogun []19:45
-!- travis-ci [~travis-ci@ec2-54-167-46-12.compute-1.amazonaws.com] has joined #shogun20:41
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/32466194020:41
-!- travis-ci [~travis-ci@ec2-54-167-46-12.compute-1.amazonaws.com] has left #shogun []20:41
-!- travis-ci [~travis-ci@ec2-54-167-46-12.compute-1.amazonaws.com] has joined #shogun21:04
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/32466194021:04
-!- travis-ci [~travis-ci@ec2-54-167-46-12.compute-1.amazonaws.com] has left #shogun []21:04
-!- braceletboy [uid256084@gateway/web/irccloud.com/x-xnapwexnihdjywyd] has quit [Quit: Connection closed for inactivity]21:30
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4066 opened by lisitsyn21:58
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4066 synchronized by lisitsyn22:16
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4066 synchronized by lisitsyn23:08
--- Log closed Thu Jan 04 00:00:05 2018
« Tuesday, 2018-01-02 Index Thursday, 2018-01-04 »

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