IRC logs of #shogun for Friday, 2018-06-01

« Thursday, 2018-05-31 Index Saturday, 2018-06-02 »
--- Log opened Fri Jun 01 00:00:31 2018
-!- iglesias [~iglesias@f119189.upc-f.chello.nl] has quit [Ping timeout: 240 seconds]02:57
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Ping timeout: 260 seconds]03:07
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun03:09
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4309 merged by vigsterkr04:36
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/e1f02a5f49f67cede1dd3ba76a4d0710d1b3a84c by vigsterkr04:36
-shogun-buildbot:#shogun- Build trusty - libshogun - viennacl #501 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/6/builds/50104:40
-shogun-buildbot:#shogun- Build nightly stretch deb #122 is complete: Failure [failed dput (failure)] - http://buildbot.shogun-toolbox.org:8080/#builders/38/builds/12205:12
-shogun-buildbot:#shogun- Build deb1 - libshogun #472 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/10/builds/47205:36
-shogun-buildbot:#shogun- Build trusty - libshogun #406 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/33/builds/40605:41
-shogun-buildbot:#shogun- Build FC23 - libshogun #506 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/27/builds/50605:41
-shogun-buildbot:#shogun- Build yakkety - libshogun #498 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/1/builds/49805:43
-shogun-buildbot:#shogun- Build zesty - libshogun #494 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/13/builds/49405:43
-shogun-buildbot:#shogun- Build xenial - libshogun #495 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/16/builds/49505:46
-shogun-buildbot:#shogun- Build deb3 - interfaces #360 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/37/builds/36006:20
-!- mikeling [uid89706@gateway/web/irccloud.com/x-cftmbciyshhazfyz] has joined #shogun08:30
-!- iglesias [~iglesias@f119189.upc-f.chello.nl] has joined #shogun08:35
-!- iglesias [~iglesias@f119189.upc-f.chello.nl] has quit [Client Quit]08:35
-!- HeikoS [~heiko@host86-128-122-106.range86-128.btcentralplus.com] has joined #shogun09:59
-!- mode/#shogun [+o HeikoS] by ChanServ09:59
-!- mikeling [uid89706@gateway/web/irccloud.com/x-cftmbciyshhazfyz] has quit [Quit: Connection closed for inactivity]10:39
@wikingyeeee! mike!10:43
@wiking:)10:43
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has joined #shogun10:49
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4321 assigned to: vigsterkr by vigsterkr10:50
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4321 unassigned vigsterkr by vigsterkr10:50
@wikingshubham808, if you happene to have any questions/problems with 4321 then wuwei could help you out as he already was involved in this task in other parts of the lib.10:57
shubham808yes sure10:58
shubham808i am still looking into %newobject10:58
@wikingkk10:59
-!- HeikoS [~heiko@host86-128-122-106.range86-128.btcentralplus.com] has quit [Quit: Leaving.]11:13
-!- durovo [~durovo@31.8f.559e.ip4.static.sl-reverse.com] has quit [Remote host closed the connection]11:31
-!- durovo [~durovo@31.8f.559e.ip4.static.sl-reverse.com] has joined #shogun11:31
@wikingwuwei, here11:51
@wiking?11:51
wuweiwiking: hi11:51
@wikingwhy is this necesarry sorry i missed the discussion11:51
@wikingauto result = wrap(cross->evaluate());11:51
@wikingauto mean = wrap(result->get<float64_t>("mean"));11:51
@wikingauto stddev = wrap(result->get<float64_t>("std_dev"));11:51
@wiking(from a metaexample of cpp)11:52
@wikingmeaning wrap(...)11:52
wuweithat's translation rule11:53
@wikingi mean in the last 2 case it's a wrap around a simple11:53
@wikingprimiteive11:53
wuweiso even trivial type is wrapped11:53
@wikingi dont get why do we need to wrap a primitive11:53
@wikingyeah but this looks super ugly11:53
@wikingand i guess it doesnt do anything11:53
@wikingin case of a primitive type11:53
@wikingright?11:53
@wiking:)11:53
wuweitype var = ...11:53
@wikingit's just a NAP11:53
@wiking*NOP11:53
wuweiThis expression is always wrapped in cop11:53
wuweicop11:53
@wikingwhere's the wrap defined?11:54
wuweicpp11:54
@wikingi mean the function itself11:54
@wiking*something* wrap(....)11:54
wuweisome.h11:54
@wikingi guess it's a template function or something11:54
@wikingk11:55
@wikingyeah i'm still not convinced about this :()11:55
@wikingi'll have a thought around it :)11:56
wuweimaybe we can do some tricks in meta example parser so that primitive types are recognized11:57
@wikingyeah something like that11:57
@wiking:)11:57
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/d8873e68f297a80bc3ce08b5e73f48a754e2bdac by vigsterkr11:59
shubham808wiking: hey12:02
@wikingho12:02
shubham808for the %newobject issue12:03
@wikingyes12:03
shubham808what we could do is like %newobject distance(...) right there in factory.i12:03
shubham808i think ?12:03
@wikingyes12:03
@wikingindeed12:03
shubham808okay12:03
@wikingfor all the factory methods12:03
@wikingwhich creates a new obj12:03
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4322 opened by shubham80812:03
@wikingthat shall be done :)12:03
-!- wiking [~wiking@huwico/staff/wiking] has left #shogun ["Leaving"]12:04
-!- wiking [~wiking@huwico/staff/wiking] has joined #shogun12:04
-!- mode/#shogun [+o wiking] by ChanServ12:04
@wikingmiju12:04
shubham808wiking: ^12:04
@wikingshubham808, yep have you tested this locally?12:04
shubham808its building okay.. but i am still doubtful12:05
@wikingk12:05
shubham808lets see what ci says12:05
@wikingokok12:05
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/432212:11
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun12:23
-!- mode/#shogun [+o HeikoS] by ChanServ12:23
@sukey[https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4283 closed by karlnapf12:30
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4287 merged by karlnapf12:30
@sukey[https://github.com/shogun-toolbox/shogun] New commit https://github.com/shogun-toolbox/shogun/commit/f14235704a0dd7c98b4569daae771378128efdd1 by karlnapf12:30
@HeikoSshubham808: cool, whats next?12:32
shubham808next i am fixing the new perceptron test12:33
shubham808then will look into jinja2 pr12:33
@HeikoSshubham808: cool!12:34
@HeikoSwiking: shubham808 will cherry pick and extend michele patch on dropping jinja from the serialization tests12:34
@HeikoSsince he will build on it for model pausing tests12:34
@wikingHeikoS, you need to take care of windows errors12:35
@HeikoSyep12:35
@HeikoSthats first step12:35
@HeikoSshubham808: ^12:35
shubham808yeah12:36
@HeikoSshubham808:  let me know if you need help12:50
shubham808sure :)12:50
shubham808HeikoS: hey13:00
-shogun-buildbot:#shogun- Build deb4 - python3 #443 is complete: Success [build successful] - http://buildbot.shogun-toolbox.org:8080/#builders/28/builds/44313:00
@HeikoSshubham808: whats up?13:00
shubham808about EXPECT_TRUE(perceptron_initialized->get_w().equals(weights))13:00
@HeikoSah yes13:00
@HeikoSyou see the difference?13:01
shubham808it is passing even for vectors that are not equal13:01
@HeikoSbetween the implementation of operator== and .equals?13:01
@HeikoSreally?13:01
shubham808mmh13:01
shubham808let me build an example for u13:01
@HeikoSshubham808: I would be quite surprised13:02
@HeikoScheck for bool SGVector<real_t>::equals(const SGVector<real_t>& other) const13:02
@HeikoSin SGVector.cpp13:02
@HeikoSthere you see the implementation13:02
shubham808https://pastebin.com/ZvpNjVA413:06
shubham808i tried giving it wrong weights values13:06
shubham808(with a lot of print statement for debugging)13:07
shubham808https://pastebin.com/2qEkKwXc13:07
@HeikoSshubham808: check13:07
@HeikoSing13:07
shubham808in the test log see from line 2513:08
@HeikoSthe code doesnt seem to match the output13:10
@HeikoSPW13:10
@HeikoSand pw13:10
@HeikoSin the output there is a PW13:10
@HeikoSand in the code there only is a pw13:10
shubham808yeah the PW is from inside Perceptron.cpp13:11
@HeikoSshubham808: ok i see13:11
@HeikoSso lets see13:11
@HeikoSfirst q I have, why doesnt it converge? :D13:11
shubham808because we gave wrong weights and set max iters to 113:12
@HeikoSw probably changes during training13:12
@HeikoScan you print both vectors just before the EXPECT_EQ call?13:12
@HeikoSuse w.display_vector()13:12
shubham808okay13:12
shubham808i see i see13:14
shubham808ofcourse its equal13:15
@HeikoSlol :D13:18
@HeikoSalways remember that copying SGVector or assigning it by value doesnt copy/replicate the underying memory13:18
shubham808yeah13:19
@HeikoSso now my q again13:19
@HeikoSif you initialize with a vector from a previously converged run13:20
@HeikoSwhy doesnt it converge after one iteration?13:20
shubham808becuase i botched it in between (to see if it fails)13:20
shubham808w.set_const(1.2)13:20
@HeikoSah ok13:21
@HeikoSsure13:21
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4320 synchronized by shubham80813:23
@HeikoSshubham808: it all works now?13:24
shubham808yeah13:24
shubham808i manually tested it all now13:24
@HeikoSshubham808: I just realised that maybe the bool converged should be a member of iterative machines13:25
@HeikoSso that a user can test whether train has converged13:25
@HeikoSbut we can do that later13:25
@HeikoSfor now the PR is good13:25
shubham808that is useful13:25
shubham808yeah13:25
@HeikoSshubham808: it is actually part of a good black box thing13:27
@HeikoSshubham808: but ok13:27
@HeikoSlet's do the test next13:27
shubham808yeah13:27
@HeikoSshubham808: let's talk here :)13:28
shubham808i am leaving for a bit13:28
shubham808lets look at the jinja pr later?13:29
@HeikoSsure13:29
shubham808bye13:29
@HeikoSsee you!13:29
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]13:29
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4322 synchronized by shubham80815:35
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has joined #shogun15:35
shubham808HeikoS: hey16:11
@HeikoSshubham808: hi16:11
shubham808we make it like %template(kernel) shogun::kernel<float64_t>; right ?16:12
@HeikoSno I think you just have to give explicit template parameters in %newobject16:15
@HeikoSthe template is already instantiated16:15
shubham808i see16:16
shubham808we wont need to do this for features factory16:18
shubham808right ?16:18
wuweihi, i just realized a problem16:22
wuweiif you add %newobject to factory, there will be leak actually, imo16:22
wuweibecause currently create_object ref the created object16:22
@HeikoSwuwei: might be16:28
@HeikoSthough there once was some convention that if you return a new object in shogun, you always REF it16:29
@HeikoSnot sure what is the state of that, needs checking16:29
@HeikoSwuwei: iirc all the meta examples leak atm ::)16:29
wuweiahh..that's too bad16:30
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4322 synchronized by shubham80816:33
-!- mikeling [uid89706@gateway/web/irccloud.com/x-icelohqzydpdbxjq] has joined #shogun16:49
shubham808HeikoS: i made an update16:58
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4317 synchronized by vinx1317:17
@wikingmikeling, hey! how is it going?17:20
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4317 synchronized by vinx1317:21
@HeikoSshubham808: let me check17:42
@HeikoSwuwei: if you manage to fix some of this, it would definitely count as one of the weekly meta example patches17:43
@HeikoSwuwei: so only need  to write a little cookbook in that case :)17:43
wuweiheiko: which one?17:44
@HeikoSwuwei: the leaks I mean18:41
wuweisure18:42
@sukey[https://github.com/shogun-toolbox/shogun] Pull Request https://github.com/shogun-toolbox/shogun/pull/4322 synchronized by shubham80818:46
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]19:00
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has joined #shogun19:00
-!- mikeling [uid89706@gateway/web/irccloud.com/x-icelohqzydpdbxjq] has quit [Quit: Connection closed for inactivity]19:09
shubham808HeikoS: state update pr is ready19:20
shubham808(the perceptron one)19:21
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]19:23
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has joined #shogun19:24
-!- shubham808 [0e8bf0fb@gateway/web/cgi-irc/kiwiirc.com/ip.14.139.240.251] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]19:34
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 240 seconds]19:43
-!- durovo3 [~durovo@31.8f.559e.ip4.static.sl-reverse.com] has joined #shogun21:08
-!- durovo [~durovo@31.8f.559e.ip4.static.sl-reverse.com] has quit [Remote host closed the connection]21:11
-!- HeikoS [~heiko@host86-128-122-106.range86-128.btcentralplus.com] has joined #shogun21:13
-!- mode/#shogun [+o HeikoS] by ChanServ21:13
-!- HeikoS [~heiko@host86-128-122-106.range86-128.btcentralplus.com] has quit [Ping timeout: 244 seconds]22:06
--- Log closed Sat Jun 02 00:00:32 2018
« Thursday, 2018-05-31 Index Saturday, 2018-06-02 »

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