--- Log opened Mon Oct 10 00:00:26 2016 | ||
-!- zxtx_ [~zxtx@sunnythinking.org] has joined #shogun | 10:55 | |
-!- Netsplit *.net <-> *.split quits: sukey, shogitter, zxtx | 10:56 | |
CaBa | hi | 12:11 |
---|---|---|
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 12:11 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 12:11 | |
@HeikoS | wiking: around? | 12:12 |
@HeikoS | cfjhallgren: jojo | 12:13 |
@wiking | HeikoS: yes | 12:14 |
@HeikoS | wiking: airbnb, see my meail | 12:28 |
@HeikoS | emaio | 12:28 |
@HeikoS | 12:28 | |
@HeikoS | wiking: should book today or so | 12:28 |
@HeikoS | did you see the listings I sent back a while ago? | 12:28 |
@wiking | yes | 12:29 |
@wiking | i saw | 12:29 |
lisitsyn | meeh | 12:31 |
@HeikoS | lisitsyn: hi | 12:33 |
lisitsyn | hey | 12:34 |
@HeikoS | lisitsyn: all good? | 12:34 |
lisitsyn | yeah | 12:34 |
lisitsyn | you? | 12:34 |
@HeikoS | all good here | 12:35 |
@HeikoS | lisitsyn: are busy are you? | 12:36 |
@HeikoS | lisitsyn: would love to have this cross-language integration testing going | 12:36 |
lisitsyn | HeikoS: a bit tired of being that kind of corporate slave :D | 12:36 |
@HeikoS | for that, we need this guy here in the modular interfaces | 12:37 |
lisitsyn | ohh it sounds very boring | 12:37 |
@HeikoS | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/lib/WrappedBasic.h | 12:38 |
@HeikoS | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/lib/WrappedObjectArray.h | 12:38 |
@HeikoS | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/lib/WrappedSGMatrix.h | 12:38 |
@HeikoS | lisitsyn: it is :) | 12:38 |
@HeikoS | but it would be increase test coverage a lot | 12:39 |
@HeikoS | which is even more boring :D | 12:39 |
lisitsyn | Im dying | 12:39 |
lisitsyn | how can you put your spare time | 12:39 |
lisitsyn | on such a b o r i n g thing | 12:39 |
@HeikoS | its the last step of the meta examples | 12:39 |
@HeikoS | then its done and I dont need to touch it again | 12:39 |
@HeikoS | testing is solved in shogun | 12:40 |
@HeikoS | so it is in anticipation of not spending time with testing anymore | 12:40 |
@HeikoS | lisitsyn: well I try again tomorrow | 12:40 |
@HeikoS | you seem to change mood ;) | 12:40 |
lisitsyn | haha | 12:40 |
lisitsyn | well | 12:40 |
@HeikoS | lisitsyn: even more fun in the admin of the foundation | 12:41 |
@HeikoS | because we did a formal mistake in the application | 12:41 |
@HeikoS | so need to send again | 12:41 |
@HeikoS | now THAT is meeeg | 12:41 |
@HeikoS | h | 12:41 |
lisitsyn | oh fuck | 12:42 |
lisitsyn | I would kill myself | 12:43 |
lisitsyn | it is the definition of boresome | 12:43 |
@HeikoS | lisitsyn: so what excites you these days? | 12:45 |
@HeikoS | used to be different! | 12:45 |
lisitsyn | hahah | 12:45 |
lisitsyn | did I say I am fancy of doing *integration testing*? :P | 12:45 |
lisitsyn | HeikoS: I would be happy to spend some time finishing get/set thingy | 12:46 |
lisitsyn | its kind of game changer | 12:47 |
@HeikoS | haha | 12:50 |
@HeikoS | hey yeah | 12:51 |
@HeikoS | that would be absolutely amazing | 12:51 |
@HeikoS | lisitsyn: finish it! | 12:51 |
@HeikoS | get rid of all the shit | 12:51 |
@HeikoS | make the plugins works | 12:51 |
@HeikoS | wokr | 12:51 |
@HeikoS | work | 12:51 |
@HeikoS | lisitsyn: so whats next step? | 12:51 |
@HeikoS | lisitsyn: all get/set patches merged? | 12:51 |
@HeikoS | whats needed? | 12:51 |
@HeikoS | lisitsyn: you know wiking and me will spend quite some time hacking shogun in two weeks | 12:52 |
@HeikoS | can we help you? | 12:52 |
lisitsyn | probably not sure | 12:52 |
lisitsyn | I'd have to check the status | 12:52 |
lisitsyn | don't remember it by heart | 12:52 |
@HeikoS | remember how the project ended? | 12:53 |
@HeikoS | I think the parameter stuff is in right? | 12:54 |
@HeikoS | it needs to be applied shogun wide now | 12:55 |
lisitsyn | kind of | 12:56 |
lisitsyn | there is a branch | 12:56 |
lisitsyn | not merged yet | 12:56 |
lisitsyn | if it is not merged it is lost | 12:56 |
@HeikoS | yeah | 12:56 |
@HeikoS | whats needed to merge it? | 12:56 |
lisitsyn | I am not even sure it is mergeable now | 12:56 |
lisitsyn | oh | 12:58 |
lisitsyn | it is! | 12:58 |
lisitsyn | that's good | 12:58 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Ping timeout: 260 seconds] | 13:25 | |
CaBa | have the examples been adopted after the GaussianKernel paramter change from 'width' to 'log_width'? | 14:13 |
CaBa | gaussian_kernel_width=ModelSelectionParameters("log_width") | 14:13 |
CaBa | gaussian_kernel_width.build_values(-math.log(2.0), 0.0, R_EXP, 1.0, 2.0) | 14:13 |
CaBa | is this example actually a useful one? | 14:13 |
-!- HeikoS [~heiko@nat-8-78.internal.eduroam.ucl.ac.uk] has joined #shogun | 14:48 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 14:48 | |
@HeikoS | lisitsyn: merged it? | 14:49 |
lisitsyn | nono not yet | 14:49 |
@HeikoS | so how can I help? | 14:49 |
lisitsyn | HeikoS: unlikely I get time before this night | 14:50 |
lisitsyn | :) | 14:50 |
@HeikoS | thats why I asked how can I help? | 14:50 |
lisitsyn | HeikoS: nah just wait :) | 14:52 |
@HeikoS | ok :) | 14:52 |
@HeikoS | I want the game to be changed :) | 14:52 |
CaBa | O.O | 14:52 |
@HeikoS | CaBa: big changes incoming ;) | 14:53 |
lisitsyn | let me just pass biyearly performance review wuhahhah | 14:53 |
@HeikoS | and more payraises ;) | 14:54 |
CaBa | HeikoS: what's happening? | 14:54 |
lisitsyn | yes that's why | 14:54 |
@HeikoS | CaBa: lisitsyn is getting a payraise ;) | 14:55 |
@HeikoS | CaBa: and some GSoC fruits to be harvested | 14:55 |
CaBa | HeikoS: what's affected? | 14:57 |
CaBa | by the way, the parallel xval is for sure functional? i'm performing a rather large grid search and see no parallelization in effect there... | 14:58 |
@HeikoS | CaBa: shoguns core architecture can be modernized then | 14:58 |
@HeikoS | CaBa: I think it works | 14:59 |
@HeikoS | you might have to set no of threads manually as detection might not work | 14:59 |
CaBa | HeikoS: how are the available threads distributed among the x-val and the machine? | 14:59 |
@HeikoS | any_sg_object->get_global_parallel()->set_num_threads(12) | 14:59 |
@HeikoS | CaBa: at the moment, we use a very crude approach | 15:00 |
CaBa | i think detection works because the SVM does use more than one thread | 15:00 |
@HeikoS | but it is effective | 15:00 |
CaBa | hmmm... my process is pretty much at ~250 % cpu permanently... | 15:00 |
@HeikoS | can you print the number of threads using the above? | 15:00 |
CaBa | on a 24 thread system | 15:00 |
@HeikoS | I mean how many shogun has set | 15:01 |
CaBa | with the corresponding getter you mean? | 15:01 |
@HeikoS | sg.GaussianKernel().get_global_parallel().get_num_threads() | 15:01 |
@HeikoS | from Pythin | 15:01 |
@HeikoS | for me returns 1 | 15:02 |
@HeikoS | even though I have 4 | 15:02 |
@HeikoS | so I need to | 15:02 |
@HeikoS | In [6]: sg.GaussianKernel().get_global_parallel().get_num_threads() | 15:02 |
@HeikoS | Out[6]: 1 | 15:02 |
@HeikoS | In [7]: sg.GaussianKernel().get_global_parallel().set_num_threads(4) | 15:02 |
@HeikoS | In [8]: sg.GaussianKernel().get_global_parallel().get_num_threads() | 15:02 |
@HeikoS | Out[8]: 4 | 15:02 |
@HeikoS | In [9]: | 15:02 |
@HeikoS | and after the xvalidation uses 4 threads | 15:02 |
@HeikoS | CaBa: let me know | 15:03 |
@HeikoS | CaBa: btw this is xvalidation that is parallelised | 15:04 |
@HeikoS | not the grid search (atm) | 15:04 |
@HeikoS | CaBa: we could work on that if you want | 15:04 |
CaBa | right... | 15:04 |
CaBa | but parallelization is over folds, not runs, right? | 15:04 |
@HeikoS | CaBa: check the code, ill send a link | 15:05 |
@HeikoS | https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/evaluation/CrossValidation.cpp#L266 | 15:06 |
@HeikoS | so we could offer multiple strategies to parallelise | 15:06 |
@HeikoS | over runs | 15:06 |
@HeikoS | over folds | 15:06 |
@HeikoS | over parameter sweeps | 15:06 |
@HeikoS | CaBa: so over folds only really helps if each fold takes some time | 15:07 |
CaBa | HeikoS: number of threads reported is 24 | 15:07 |
@HeikoS | but then helps a lot | 15:07 |
@HeikoS | ok | 15:07 |
@HeikoS | how many folds are you doing? | 15:07 |
@HeikoS | how many runs? | 15:07 |
@HeikoS | how many points does your grid have? | 15:07 |
@HeikoS | 24 folds is unlikely I guess ;) | 15:07 |
@HeikoS | so we could add parallelisation over runs as well? | 15:08 |
@HeikoS | or do it in the grid search | 15:08 |
@HeikoS | whatever suits you | 15:08 |
@HeikoS | you willing to send a patch? | 15:08 |
@HeikoS | I can help | 15:08 |
CaBa | HeikoS: during model selection i use 1 run, 10 folds | 15:08 |
@HeikoS | but the code in the link should give a good idea of how it works | 15:08 |
@HeikoS | ok, so you cant fully exploit your setupo | 15:09 |
CaBa | so the folds won't run in parallel? | 15:09 |
@HeikoS | they do | 15:09 |
@HeikoS | check the link | 15:09 |
@HeikoS | also tested etc | 15:10 |
CaBa | strange | 15:10 |
@HeikoS | oh and only if machine is not locked | 15:10 |
@HeikoS | can you check whether yours is locked? | 15:10 |
CaBa | HeikoS: CCrossValidation() is initialized with autolock=true | 15:12 |
CaBa | HeikoS: but i dont manually lock the machine anywhere | 15:12 |
@HeikoS | can you set it to false and try? | 15:15 |
@HeikoS | do you use a kernel machine there? | 15:16 |
CaBa | yes | 15:16 |
CaBa | svm | 15:16 |
@HeikoS | I see | 15:16 |
@HeikoS | try disabling the autolock | 15:16 |
@HeikoS | just for debugging | 15:16 |
@HeikoS | then cpu should go up | 15:16 |
@HeikoS | and then I can add parallel xvalidation for precomputed kernel matrices | 15:17 |
@HeikoS | shouldnt be too hard | 15:17 |
@HeikoS | and its good to have a use case from your side | 15:17 |
CaBa | HeikoS: indeed now i seem to have a higher CPU usage | 15:20 |
@HeikoS | CaBa: ok! | 15:21 |
CaBa | HeikoS: ~660% instead of ~230% | 15:21 |
@HeikoS | that's it then | 15:21 |
@HeikoS | ironically, it might slower after all | 15:21 |
@HeikoS | since kernel matrix is not precomputed now | 15:21 |
@HeikoS | do you notice? | 15:21 |
@HeikoS | but we can add that feature back in | 15:21 |
@HeikoS | let me know | 15:22 |
@HeikoS | gotta run now, but back later | 15:22 |
CaBa | can't say, the runs take me half a day | 15:22 |
CaBa | i'm really quite frustrated with that to be honest :( | 15:22 |
CaBa | ok, cool cu later | 15:22 |
-!- HeikoS [~heiko@nat-8-78.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 240 seconds] | 15:48 | |
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun | 15:57 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 15:57 | |
-!- mizari [~mizari@95-174-213-100.nts.su] has joined #shogun | 16:20 | |
CaBa | is there an interface to get the kernel matrix out of a machine? | 17:01 |
-!- HeikoS1 [~heiko@nat-8-78.internal.eduroam.ucl.ac.uk] has joined #shogun | 17:36 | |
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 250 seconds] | 17:38 | |
Saurabh7 | yello | 18:17 |
Saurabh7 | CaBa: ckernel->get_kernel_matrix i think .. | 18:18 |
HeikoS1 | Caba yes | 18:37 |
HeikoS1 | thats it ^ | 18:37 |
CaBa | thanks | 18:37 |
Saurabh7 | HeikoS1: hey ! | 18:37 |
Saurabh7 | how are you | 18:37 |
CaBa | HeikoS1: is that a minor change to make the parallel xval work on locked machines? | 18:38 |
Saurabh7 | ah yes that was left unfinished | 18:38 |
Saurabh7 | I can attempt to do something this week :) | 18:39 |
-!- HeikoS1 [~heiko@nat-8-78.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 260 seconds] | 18:41 | |
Saurabh7 | I had it working for locked iirc, but it was hackish, so HeikoS made me remove it :p | 18:45 |
CaBa | ^^ | 18:45 |
-!- HeikoS [~heiko@nat-8-78.internal.eduroam.ucl.ac.uk] has joined #shogun | 18:58 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 18:58 | |
-!- HeikoS [~heiko@nat-8-78.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 260 seconds] | 19:02 | |
-!- c4goldsw [8a33f9de@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.249.222] has joined #shogun | 19:57 | |
c4goldsw | Hey wiking, you around? | 19:57 |
-!- c4goldsw [8a33f9de@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.249.222] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 20:09 | |
-!- mizari [~mizari@95-174-213-100.nts.su] has quit [Quit: Leaving] | 20:18 | |
-!- GandalfTheWizard [~Emma@112.10.170.100] has joined #shogun | 20:20 | |
-!- GandalfTheWizard [~Emma@112.10.170.100] has quit [Max SendQ exceeded] | 20:20 | |
-!- GandalfTheWizard [~Emma@112.10.170.100] has joined #shogun | 20:21 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 20:24 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 20:24 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Client Quit] | 20:25 | |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 20:25 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 20:25 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has joined #shogun | 20:37 | |
c4goldsw | Hey HeikoS | 20:45 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Quit: Leaving.] | 20:47 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 21:05 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has joined #shogun | 21:15 | |
-!- GandalfTheWizard [~Emma@112.10.170.100] has quit [Ping timeout: 260 seconds] | 21:32 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 21:40 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has joined #shogun | 21:41 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 22:19 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has joined #shogun | 22:21 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 22:35 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has joined #shogun | 22:36 | |
-!- c4goldsw [8a33fa12@gateway/web/cgi-irc/kiwiirc.com/ip.138.51.250.18] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 23:10 | |
--- Log closed Tue Oct 11 00:00:27 2016 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!