--- Log opened Fri May 27 00:00:13 2016 | ||
-!- Saurabh7- [Saurabh7@gateway/shell/panicbnc/x-tkhunxevfgkzzhnq] has quit [Read error: Connection reset by peer] | 00:39 | |
-!- Saurabh7 [Saurabh7@gateway/shell/panicbnc/x-zjbrorlyietxjdsm] has joined #shogun | 00:42 | |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 272 seconds] | 01:15 | |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 01:23 | |
-!- mode/#shogun [+o besser82] by ChanServ | 01:23 | |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 260 seconds] | 01:39 | |
shogun-buildbot | build #6 of clang - thread analysis is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/clang%20-%20thread%20analysis/builds/6 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 03:44 |
---|---|---|
shogun-buildbot | build #5 of clang - undefined behaviour analysis is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/clang%20-%20undefined%20behaviour%20analysis/builds/5 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 03:47 |
-!- OXPHOS [9d8b131c@gateway/web/freenode/ip.157.139.19.28] has quit [Quit: Page closed] | 03:49 | |
-!- sonne|osx_ [~sonne@x4db38903.dyn.telefonica.de] has joined #shogun | 05:54 | |
-!- sonne|osx [~sonne@x4e33d889.dyn.telefonica.de] has quit [Ping timeout: 272 seconds] | 05:56 | |
-!- sonne|osx_ is now known as sonne|osx | 05:56 | |
shogun-buildbot | build #5 of memleak - valgrind is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/memleak%20-%20valgrind/builds/5 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 06:16 |
shogun-buildbot | build #1008 of nightly_none is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_none/builds/1008 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 06:27 |
shogun-buildbot | build #1137 of nightly_default is complete: Failure [failed test notebooks] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/1137 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 07:41 |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has joined #shogun | 11:50 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 11:50 | |
@wiking | HeikoS: hey ehy | 11:50 |
@HeikoS | wiking: jo | 11:51 |
@HeikoS | all good? | 11:51 |
@wiking | standups | 11:52 |
@wiking | :) | 11:52 |
@wiking | i'm missing those | 11:52 |
@wiking | https://en.wikipedia.org/wiki/Gold_%28linker%29 | 11:53 |
lisitsyn | ok let me start | 11:53 |
lisitsyn | NDA NDA NDA NDA | 11:53 |
lisitsyn | NDA NDA NDA | 11:53 |
lisitsyn | going to finish with NDA | 11:53 |
@HeikoS | wiking: | 11:53 |
@HeikoS | ping everyone! | 11:53 |
@wiking | ???????? | 11:53 |
@wiking | lisitsyn: nda? | 11:53 |
@wiking | HeikoS: https://en.wikipedia.org/wiki/Gold_%28linker%29 | 11:54 |
@HeikoS | arianepaola, Saurabh7, arianepaola stand up reminder :) | 11:54 |
lisitsyn | wiking: yes that's my standup | 11:54 |
@HeikoS | wiking: keepo in mind time zone | 11:54 |
lisitsyn | wiking: I told you what I am up to | 11:54 |
@HeikoS | I just started | 11:54 |
lisitsyn | :D | 11:54 |
@HeikoS | wiking: tried it? | 11:54 |
@wiking | HeikoS: yes | 11:55 |
lisitsyn | what's gold and why | 11:55 |
lisitsyn | faster? | 11:56 |
@HeikoS | wiking: and? | 11:56 |
@wiking | HeikoS: just running the time for make | 11:57 |
@HeikoS | wiking: I see | 11:57 |
@HeikoS | wiking: btw ever tried ninja on shogun? | 11:57 |
@HeikoS | also is a bit faster, especially when re-compiling | 11:57 |
@HeikoS | arianepaola told me abou tit | 11:57 |
@wiking | ninja? | 11:57 |
@wiking | what the fuck is ninja? | 11:57 |
lisitsyn | some alternative to make | 11:58 |
lisitsyn | cmake can generate that | 11:58 |
@wiking | aaaah | 11:58 |
@wiking | i really dont care about make's speed | 11:59 |
@wiking | more linker speed | 11:59 |
@wiking | unit test generation is really a pain now | 11:59 |
@wiking | (trying to fix memleaks) | 12:00 |
lisitsyn | oh btw | 12:00 |
lisitsyn | can I help? | 12:00 |
@wiking | lisitsyn: memleaks :) | 12:00 |
lisitsyn | yes | 12:00 |
@wiking | i've fixed some | 12:00 |
lisitsyn | 'some' | 12:00 |
lisitsyn | or some? | 12:00 |
@wiking | there's a really nasty one | 12:00 |
lisitsyn | what was wrong? | 12:00 |
@wiking | lisitsyn: look at the generated one | 12:00 |
@wiking | plz | 12:00 |
@wiking | why do they leak | 12:00 |
@wiking | or why does valgrind say that they leak | 12:01 |
lisitsyn | wiking: ok | 12:01 |
@wiking | lisitsyn: | 12:01 |
@wiking | http://buildbot.shogun-toolbox.org/memcheck/20160527-0100.html | 12:01 |
lisitsyn | wiking: well I'd expect off-by-one in ref counter | 12:02 |
lisitsyn | let me check | 12:02 |
@wiking | thnx | 12:02 |
@HeikoS | wiking: so is the gold linker faster? | 12:07 |
@HeikoS | wiking: I would not worry too much about the generated examples right now | 12:07 |
@HeikoS | wiking: even if they produce many leaks each, it is a single fix that fixes all of them | 12:08 |
@HeikoS | or do you disagree? | 12:08 |
@HeikoS | valgrind told me its the new operator on CSGOBjects that is not freed | 12:08 |
@wiking | testing gold | 12:10 |
@wiking | it should be faster | 12:10 |
@wiking | :) | 12:10 |
-!- HeikoS_mobile [~Mutter@nat-195-25.internal.eduroam.ucl.ac.uk] has joined #shogun | 12:11 | |
lisitsyn | mobile! | 12:11 |
HeikoS_mobile | Yes!! | 12:11 |
HeikoS_mobile | lisitsyn: coming to london soon? | 12:11 |
lisitsyn | HeikoS_mobile: not very likely | 12:12 |
HeikoS_mobile | Nooooo | 12:12 |
HeikoS_mobile | Just weekend? | 12:12 |
lisitsyn | HeikoS_mobile: well weekend is likely to be possible %) | 12:13 |
lisitsyn | Could NOT find PythonLibs: Found unsuitable version "2.7.10", but required | 12:14 |
lisitsyn | is exact version "2.7.11" | 12:14 |
lisitsyn | why do I get this? | 12:14 |
lisitsyn | wiking: why do we require 2.7.11? :) | 12:15 |
@wiking | ah | 12:16 |
@wiking | lisitsyn: this is because you have 2 pythons | 12:16 |
@wiking | and it'll mix and match and it'll fuck up your python modular interface | 12:16 |
-!- HeikoS_mobile [~Mutter@nat-195-25.internal.eduroam.ucl.ac.uk] has quit [Read error: Connection reset by peer] | 12:17 | |
@wiking | it detects python 2.7.11 interpreter (probably /usr/local/bin/python or /opt/local/bin/python) | 12:17 |
@wiking | but then it detects the python lib to be in /usr/lib/ | 12:17 |
@wiking | which is the system python | 12:17 |
@wiking | not the interpreter | 12:17 |
lisitsyn | wiking: hmm not really I have installed python | 12:18 |
lisitsyn | and gave it a prefix | 12:18 |
@wiking | lisitsyn: can you pastebin your full cmake plz | 12:19 |
lisitsyn | wiking: http://pastebin.com/UEdnxMX6 | 12:20 |
lisitsyn | this looks strange | 12:20 |
@wiking | ooooooo | 12:20 |
@wiking | how is this possible? :D | 12:20 |
@wiking | -- Found PythonInterp: /Users/lisitsyn/Prefix/bin/python (found version "2.7.11") | 12:20 |
@wiking | is exact version "2.7.11" (found /Users/lisitsyn/Prefix/lib/libpython2.7.a) | 12:20 |
@wiking | but that supposed to be 2.7.10? | 12:20 |
lisitsyn | ehmm ok may be I need to reinstall it | 12:21 |
@wiking | HeikoS: hahahahah | 12:25 |
@wiking | gold is slower | 12:26 |
@wiking | waaaaaaaaaaaaay slower | 12:26 |
@wiking | HeikoS: the problem with linking is the generated unit tests actually | 12:26 |
@HeikoS | wiking: hehew | 12:29 |
@HeikoS | wiking: so I think splitting the tests woul dbe good | 12:29 |
@wiking | those big serialization tests | 12:29 |
@HeikoS | have multiple executables | 12:29 |
@wiking | no | 12:29 |
@HeikoS | yeah | 12:29 |
@wiking | that is a fucking pain | 12:29 |
@wiking | we dont want that | 12:29 |
@wiking | just remove the serialization stuff | 12:29 |
@HeikoS | the thing is | 12:29 |
@wiking | once cereal is in | 12:29 |
@HeikoS | if I develop code | 12:29 |
@HeikoS | I usually do test driven | 12:29 |
@HeikoS | but I have to compile ALL unit tests every time | 12:30 |
@HeikoS | annoyintg | 12:30 |
@HeikoS | dont want that | 12:30 |
@HeikoS | I just want mine | 12:30 |
@HeikoS | currently not possible | 12:30 |
@wiking | but it'll compile only that | 12:30 |
@HeikoS | well | 12:30 |
@wiking | just linking takes a while | 12:30 |
@HeikoS | I then often do a brnach change to fix something else | 12:30 |
@HeikoS | and switch from say Debug to Release | 12:30 |
@HeikoS | to test performanc | 12:30 |
@HeikoS | then go back | 12:30 |
@wiking | but there's nothing you can do about that | 12:30 |
@HeikoS | and have to re-compile | 12:30 |
@wiking | yeah but man | 12:30 |
@wiking | you are asking about something | 12:30 |
@wiking | that is impossible to solve normally | 12:31 |
@wiking | use 2 sources | 12:31 |
@HeikoS | I was thinking after plugins arrived | 12:31 |
@HeikoS | and we have a dependency structure | 12:31 |
@HeikoS | so that I can just re-compile stuff I need/want | 12:31 |
@HeikoS | we can apply that to the tests somehow | 12:31 |
@HeikoS | but well | 12:32 |
@HeikoS | my dreams, you know ;) | 12:32 |
@HeikoS | wiking: but ok lets remove serialization then | 12:32 |
@HeikoS | after cereal is there | 12:32 |
@HeikoS | should be much faster to start with | 12:32 |
@wiking | because now we test and generate code | 12:33 |
@HeikoS | yeah I know | 12:33 |
@wiking | for every class | 12:33 |
@HeikoS | maybe we can give cmake a regexp | 12:33 |
@wiking | i mean the fucking unit test exec | 12:33 |
@wiking | is 161 megs | 12:33 |
@HeikoS | for the folders to include unit test sources from | 12:33 |
@HeikoS | haha | 12:33 |
@HeikoS | wiking: actually wouldnt that work? | 12:33 |
@wiking | that's why linking takes forever | 12:33 |
@wiking | ? | 12:33 |
@wiking | just remove the generated code and done | 12:33 |
@HeikoS | you think? | 12:34 |
@HeikoS | still compiling them takes loing | 12:34 |
@HeikoS | on my machine | 12:34 |
@HeikoS | and have to re-compile when doing branch switches often | 12:34 |
@wiking | HeikoS: then you dont have ccache normally setup | 12:34 |
@HeikoS | so if we have a cmake option | 12:34 |
@HeikoS | wiking: I do | 12:34 |
@HeikoS | its if too many things change at once | 12:34 |
@wiking | ccache -s | 12:34 |
@wiking | what's the outptu of that? | 12:35 |
@HeikoS | not found | 12:35 |
@HeikoS | ah | 12:35 |
@HeikoS | space | 12:35 |
@HeikoS | sorry | 12:35 |
@HeikoS | cache size 23.4 Gbyte | 12:35 |
@wiking | and how big is your ccache? | 12:35 |
@wiking | max cache size | 12:35 |
@HeikoS | 61Mb | 12:36 |
@wiking | you mean gbyte | 12:36 |
@wiking | i guess | 12:36 |
@HeikoS | nope | 12:36 |
@wiking | can you fucking copy paste the whole thing | 12:36 |
@HeikoS | files in cache 33500 | 12:36 |
@HeikoS | cache size 23.4 Gbytes | 12:36 |
@HeikoS | max cache size 61.0 Mbytes | 12:36 |
@wiking | plz | 12:36 |
@wiking | hahahaha | 12:36 |
@wiking | man | 12:36 |
@HeikoS | sec | 12:36 |
@wiking | is this you think | 12:36 |
@HeikoS | wiking: I never touched this at all | 12:36 |
@wiking | a normally setup | 12:36 |
@wiking | ccache? | 12:36 |
@wiking | :D | 12:36 |
@HeikoS | no idea, using default settings | 12:36 |
@wiking | yeah | 12:36 |
@wiking | hence i said | 12:36 |
@HeikoS | ok let me try then | 12:36 |
@wiking | 12:32 <@wiking> HeikoS: then you dont have ccache normally setup | 12:36 |
@HeikoS | can you shed some light on good options? | 12:37 |
@wiking | so first of all | 12:37 |
@wiking | clear the fucking cache | 12:37 |
@HeikoS | maybe we can put that in some readme | 12:37 |
@HeikoS | did before | 12:37 |
@wiking | HeikoS: maybe people start RTFM | 12:37 |
@HeikoS | didnt help | 12:37 |
@wiking | no? | 12:37 |
@wiking | :) | 12:37 |
@wiking | i mean fuck | 12:37 |
@wiking | we are not like | 12:37 |
@wiking | "how to use your computer" | 12:37 |
@wiking | ccache -C -z | 12:37 |
@wiking | do this first | 12:37 |
@HeikoS | wiking: maybe think a sec | 12:37 |
@HeikoS | if I get this wrong, how likely is that users get this wrong? | 12:38 |
@wiking | and then do | 12:38 |
@wiking | ccache -M10G | 12:38 |
@wiking | and after the first compile | 12:38 |
@wiking | you'll have it good | 12:38 |
@HeikoS | ok trying | 12:38 |
lisitsyn | hey but people don't have to compile that | 12:38 |
@wiking | but | 12:38 |
@HeikoS | thanks! | 12:38 |
@wiking | if you keep switching | 12:38 |
@wiking | between release and debug | 12:38 |
@wiking | that's gonna be always a pain | 12:38 |
@HeikoS | wiking: uh | 12:38 |
@HeikoS | thats when it happens | 12:38 |
@wiking | because that is totally different | 12:38 |
@wiking | compiler flags | 12:38 |
@HeikoS | I know | 12:38 |
@wiking | but again | 12:39 |
@HeikoS | this is why I suggest the option | 12:39 |
@wiking | having 10G | 12:39 |
@wiking | hopefully will cache you both | 12:39 |
@HeikoS | cmake option: regexp of directories to include for unit testing | 12:39 |
@wiking | that option is bull fucking shit | 12:39 |
@wiking | i'm sorry | 12:39 |
@wiking | you are asking for thing | 12:39 |
@HeikoS | why? | 12:39 |
@wiking | that is not normal | 12:39 |
@wiking | HeikoS: do a buildbot | 12:39 |
@wiking | it compiles | 12:39 |
@wiking | release | 12:39 |
@wiking | and has ccache | 12:39 |
@wiking | but again | 12:39 |
@wiking | first do what i said | 12:39 |
@wiking | test it | 12:39 |
@HeikoS | I did | 12:39 |
@wiking | and you'll see | 12:39 |
@HeikoS | whats wrong with the option? | 12:39 |
@wiking | which option? | 12:39 |
@wiking | the regexp blalbla | 12:40 |
@wiking | come ooooon | 12:40 |
@HeikoS | yes | 12:40 |
@wiking | oh coooooooooomeeeeeee oooooooooooooon | 12:40 |
@HeikoS | like "unit/classifier/*;unit/kernel*" | 12:40 |
@wiking | yes yes | 12:40 |
@wiking | lala | 12:40 |
@HeikoS | and then only those are compiled | 12:40 |
lisitsyn | why? | 12:40 |
@wiking | talk to the hand | 12:40 |
@HeikoS | because it saves me time | 12:40 |
lisitsyn | but you don't know what tests could you break | 12:40 |
@wiking | make does | 12:41 |
@HeikoS | lisitsyn: yes but I dont care since I develop a new feature that only affects my newly written tests | 12:41 |
@wiking | what it knows what it needs to do | 12:41 |
@wiking | if you are trying to be smarter | 12:41 |
@wiking | than make | 12:41 |
@wiking | have fun | 12:41 |
@wiking | i dont see the point of reinventing the wheel there | 12:41 |
@HeikoS | wiking: you havent told me *why* is is a bad idea in your eyes | 12:41 |
@wiking | just did | 12:41 |
@HeikoS | I am not trying to be smarter just trying to understand | 12:42 |
@wiking | you are tyring to solve things | 12:42 |
@wiking | that are solved | 12:42 |
@wiking | by tools | 12:42 |
@wiking | already | 12:42 |
@wiking | and if that's not good enough for you | 12:42 |
@wiking | you are gonna have to reinvent the whole thing | 12:42 |
@wiking | i'm not standing in your way | 12:42 |
@wiking | but | 12:42 |
@wiking | experiment with that | 12:42 |
@wiking | first in a dummy library | 12:42 |
@wiking | and make it feasible | 12:42 |
@wiking | and running on that w/o errors | 12:42 |
@wiking | if that's good | 12:42 |
@wiking | than MAAYBE | 12:42 |
@wiking | maaaybe it's a good idea to have it in shogun | 12:42 |
@wiking | again | 12:42 |
@wiking | you are trying to solve what's make is for | 12:42 |
@HeikoS | I dont see why it is hard | 12:43 |
@wiking | or ninja | 12:43 |
@wiking | it's not about being ahrd | 12:43 |
@HeikoS | just about including things in cmakelist in the unit test dir | 12:43 |
@wiking | it's about introducing | 12:43 |
@wiking | fucking errors | 12:43 |
@wiking | possible source of errors | 12:43 |
@wiking | look at the current code | 12:43 |
@wiking | even though | 12:43 |
@HeikoS | what type of error? | 12:43 |
@wiking | we have tets for every fucking things | 12:43 |
@wiking | we still manage | 12:43 |
@wiking | to have leaking shit every time | 12:43 |
@wiking | i look at the code 2 months later | 12:43 |
@HeikoS | you maybe dont get what I want from that | 12:43 |
@wiking | i know what you want | 12:43 |
@HeikoS | I just want to use this locally | 12:43 |
@HeikoS | no build will have the option set on | 12:44 |
@wiking | yeah | 12:44 |
@wiking | but if you set the option to do this | 12:44 |
@wiking | then you possible will have somebody use it | 12:44 |
@wiking | and then if it's used | 12:44 |
@wiking | it'll possible get some code in | 12:44 |
@wiking | somewhere | 12:44 |
@wiking | i mean for fuck sakes | 12:44 |
@wiking | we managed to merge code in | 12:44 |
@wiking | that fucking leaks | 12:44 |
@wiking | in the obvious way | 12:44 |
@wiking | do you think if you allow this to have locally | 12:44 |
@wiking | the same thing wont happen | 12:44 |
@HeikoS | that is very different | 12:44 |
@wiking | yeah very but still the same | 12:44 |
@wiking | it's human error | 12:45 |
@HeikoS | I dont agree | 12:45 |
@wiking | ok good | 12:45 |
@HeikoS | very different kind of problem | 12:45 |
@wiking | then we dont agree | 12:45 |
@wiking | first | 12:45 |
@HeikoS | sure fine with that :D | 12:45 |
@wiking | start using ccache | 12:45 |
@wiking | in a normal way | 12:45 |
@wiking | not with max size of 64 megabytes | 12:45 |
@wiking | when that really doesn't work for you | 12:45 |
@HeikoS | wiking: do you know why this is default size? | 12:45 |
@wiking | it's not | 12:45 |
@HeikoS | mmh | 12:45 |
lisitsyn | the default size is 1GB | 12:45 |
lisitsyn | iirc | 12:45 |
@HeikoS | weird | 12:45 |
@wiking | default is usually 1 or 5 | 12:45 |
lisitsyn | in ubuntu | 12:45 |
@wiking | G | 12:45 |
@wiking | not M | 12:45 |
@HeikoS | okok | 12:45 |
@HeikoS | lets see | 12:45 |
@HeikoS | hopefully that solves it | 12:46 |
@wiking | since we eliminated a lot of -D compiler flags | 12:46 |
@HeikoS | wiking: you know just really trying to play a bit with the idea | 12:46 |
@wiking | it works quite nicely | 12:46 |
@HeikoS | not saying we should do this thing | 12:46 |
@HeikoS | just thinking loud | 12:46 |
@wiking | see for example the performance of the qemu based aarch64 bot | 12:46 |
@HeikoS | kk | 12:46 |
-!- lambday [8028b10a@gateway/web/freenode/ip.128.40.177.10] has joined #shogun | 12:46 | |
-!- mode/#shogun [+o lambday] by ChanServ | 12:47 | |
@wiking | 2 hours something | 12:47 |
@wiking | down to 40 mins | 12:47 |
@wiking | that's ccache for you | 12:47 |
@HeikoS | wiking: yeah I see this here as well | 12:47 |
@wiking | well not likely | 12:47 |
@wiking | with 64 megs | 12:47 |
@wiking | :) | 12:47 |
@wiking | sorry i couldn't stand it | 12:47 |
@wiking | there are soooo many other things to fix | 12:47 |
@HeikoS | wiking: any ideas on the state of cereal btw | 12:48 |
@wiking | HeikoS: it's not detectable | 12:48 |
@wiking | :DDD | 12:48 |
@wiking | that's where we've managed to get | 12:48 |
@HeikoS | droping serialization code would reduce binary a lot I guess ;) | 12:48 |
@HeikoS | haha | 12:48 |
@HeikoS | ah man | 12:48 |
@HeikoS | this all should have happened last month I guess | 12:48 |
@wiking | so hopefully today | 12:48 |
@HeikoS | yeah | 12:48 |
@wiking | we'll start getting serialization | 12:48 |
@HeikoS | wiking: btw | 12:48 |
@HeikoS | you know the structure of the shogun structures | 12:49 |
@HeikoS | container, string, basic | 12:49 |
@HeikoS | and in these guys you can put float, SGVector, SGMatrix, SGString etc | 12:49 |
@HeikoS | so too many possibilities | 12:49 |
@HeikoS | who needs strings of matrices? | 12:49 |
@HeikoS | so I suggest to simplify that | 12:50 |
@wiking | nooooo | 12:50 |
@wiking | what do you do with | 12:50 |
@wiking | categoricals? | 12:50 |
@HeikoS | ? | 12:50 |
@wiking | how do you represent categoricals | 12:50 |
@wiking | like | 12:50 |
@wiking | feature: "low", "medium", "high" | 12:51 |
@HeikoS | kind of dont see the connection to what I just said | 12:51 |
@HeikoS | just want to get rid of all these combinations that we never use | 12:52 |
@HeikoS | string of matrices e.g. | 12:52 |
@HeikoS | matrix of matrices | 12:52 |
@wiking | but again | 12:52 |
@wiking | let's say i have | 12:52 |
@wiking | a Features | 12:52 |
@wiking | where it only has strings | 12:52 |
@wiking | how do you represent that | 12:52 |
@wiking | ? | 12:52 |
@wiking | i mean Features = SGMatrix | 12:52 |
@wiking | sorry | 12:52 |
@wiking | DenseFeatures = SGMatrix | 12:52 |
@wiking | right/ | 12:52 |
@HeikoS | yeah | 12:53 |
@HeikoS | string features still works | 12:53 |
@HeikoS | just only strings of basic types | 12:53 |
@HeikoS | not of SGVectors type | 12:53 |
@HeikoS | same with array | 12:53 |
@HeikoS | so categoricals are just represented using ints | 12:55 |
@HeikoS | and we can have strings of that, and matrices | 12:56 |
@HeikoS | and vectors | 12:56 |
@HeikoS | what I want to avoid is this collection of SG* types | 12:56 |
@HeikoS | because that is never used anyways | 12:56 |
@HeikoS | and it accounds for most of the code in current parameter fraemwork | 12:57 |
@HeikoS | wiking: ^ | 12:57 |
-!- sonne|osx [~sonne@x4db38903.dyn.telefonica.de] has quit [Quit: sonne|osx] | 13:17 | |
-!- leagoetz [~leagoetz@nat-247-179.internal.eduroam.ucl.ac.uk] has joined #shogun | 13:34 | |
-!- HeikoS_mobile_ [~Mutter@nat-195-25.internal.eduroam.ucl.ac.uk] has joined #shogun | 13:51 | |
-!- HeikoS_mobile__ [~Mutter@82-132-239-80.dab.02.net] has joined #shogun | 13:51 | |
-!- HeikoS_mobile_ [~Mutter@nat-195-25.internal.eduroam.ucl.ac.uk] has quit [Read error: Connection reset by peer] | 13:52 | |
-!- HeikoS_mobile__ [~Mutter@82-132-239-80.dab.02.net] has quit [Remote host closed the connection] | 13:54 | |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 244 seconds] | 14:25 | |
shogun-buildbot | build #2877 of bsd1 - libshogun is complete: Failure [failed configure] Build details are at http://buildbot.shogun-toolbox.org/builders/bsd1%20-%20libshogun/builds/2877 blamelist: Viktor Gal <viktor.gal@maeth.com> | 14:50 |
shogun-buildbot | build #15 of xenial - libshogun is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/xenial%20-%20libshogun/builds/15 blamelist: Viktor Gal <viktor.gal@maeth.com> | 14:51 |
@wiking | shogun-buildbot: force build --branch=develop 'memleak - valgrind' | 14:54 |
shogun-buildbot | The build has been queued, I'll give a shout when it starts | 14:54 |
arianepaola | hello everyone | 15:09 |
leagoetz | hey | 15:09 |
@wiking | hi | 15:10 |
arianepaola | wikining: I am continuing to work on the unit tests FisherLDA_unittest.cc and LDA_unittest.cc | 15:16 |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has joined #shogun | 15:17 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 15:17 | |
@wiking | arianepaola: the fisher one is quite simple | 15:17 |
@wiking | it's just a -1* | 15:17 |
@wiking | LDA is something that needs to be debugged | 15:18 |
@wiking | imo the FisherLDA unit tests should be changed | 15:18 |
@wiking | as it's an eigen vector with opposite sign | 15:18 |
@wiking | the LDA svd stuff | 15:18 |
arianepaola | yes value * (-1) | 15:18 |
@wiking | there needs to be fully debugged | 15:18 |
@wiking | where is the big change | 15:19 |
@wiking | once we know it | 15:19 |
@wiking | we can consult ggael or ChriSopht on #eigen | 15:19 |
@wiking | they are pretty helpful | 15:19 |
@wiking | they fixed a bug in develop eigen within a day i reported | 15:19 |
@wiking | i just didn't have time to debug | 15:20 |
@wiking | where and what is happening actually | 15:20 |
@wiking | within the LDA SVD part | 15:20 |
@wiking | so if you could trace where the first time a matrix' value changes | 15:20 |
@wiking | that would be super helpful | 15:20 |
@wiking | either gdb/lldb or just simply std::cout << matrix << std::endl; | 15:20 |
arianepaola | yes I will also get the latest eigen then | 15:20 |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 15:26 | |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has joined #shogun | 15:28 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 15:28 | |
shogun-buildbot | build #6 forced | 15:29 |
shogun-buildbot | I'll give a shout when the build finishes | 15:29 |
Saurabh7 | HeikoS: hi ! | 15:35 |
@wiking | Saurabh7: daily :) | 15:36 |
Saurabh7 | wiking: hi :) | 15:36 |
Saurabh7 | got parallel xval working | 15:36 |
Saurabh7 | with kernel svm for now | 15:36 |
@wiking | plan for next? | 15:37 |
Saurabh7 | and denseFeatures | 15:37 |
@wiking | any blocking issue? | 15:37 |
Saurabh7 | I will push in branch i think | 15:37 |
Saurabh7 | I need advice how to make it general | 15:37 |
@wiking | if you share the code | 15:37 |
@wiking | then i can help | 15:37 |
@wiking | and i can commet | 15:37 |
@wiking | *comment | 15:37 |
Saurabh7 | ok doing | 15:38 |
@wiking | thnx | 15:38 |
@wiking | do you have a feature branch? | 15:38 |
Saurabh7 | no | 15:38 |
@wiking | mmm we could create one,... HeikoS ? | 15:38 |
Saurabh7 | i will do it in normal one for now lets see | 15:38 |
@wiking | k | 15:38 |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 260 seconds] | 15:39 | |
arianepaola | hi Saurabh7 | 15:39 |
shogun-buildbot | build #6 of FC23 - libshogun - aarch64 is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/FC23%20-%20libshogun%20-%20aarch64/builds/6 blamelist: Viktor Gal <viktor.gal@maeth.com> | 15:42 |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has joined #shogun | 15:44 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 15:44 | |
Saurabh7 | arianepaola: hi! | 15:45 |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 240 seconds] | 15:53 | |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has joined #shogun | 16:00 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:00 | |
@HeikoS | Saurabh7: hi | 16:07 |
@HeikoS | how are things going? | 16:07 |
Saurabh7 | HeikoS: hi | 16:07 |
Saurabh7 | I got xval working in parallel now | 16:07 |
@HeikoS | Saurabh7: great! | 16:07 |
Saurabh7 | for kernel svm | 16:07 |
@HeikoS | send a PR | 16:07 |
Saurabh7 | not in general | 16:08 |
@HeikoS | even if not done, just mention that in the description | 16:08 |
@HeikoS | thats fine | 16:08 |
@HeikoS | wanna look at it | 16:08 |
Saurabh7 | need help i am pshing to branch for u to see | 16:08 |
@HeikoS | Saurabh7: you can just push to your fork and then send a pr | 16:08 |
Saurabh7 | yup | 16:08 |
@HeikoS | Saurabh7: for now, you can send that against develop | 16:09 |
-!- HeikoS_mobile__ [~Mutter@nat-195-25.internal.eduroam.ucl.ac.uk] has joined #shogun | 16:10 | |
-!- HeikoS_mobile__ [~Mutter@nat-195-25.internal.eduroam.ucl.ac.uk] has quit [Remote host closed the connection] | 16:14 | |
-!- HeikoS [~heiko@nat-196-45.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 276 seconds] | 16:14 | |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 16:24 | |
-!- mode/#shogun [+o besser82] by ChanServ | 16:24 | |
-!- OXPHOS [9d8b131c@gateway/web/freenode/ip.157.139.19.28] has joined #shogun | 16:25 | |
-!- leagoetz [~leagoetz@nat-247-179.internal.eduroam.ucl.ac.uk] has quit [] | 16:38 | |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 252 seconds] | 16:57 | |
@wiking | OXPHOS: yo | 16:59 |
@wiking | OXPHOS: as said time for squash and get the cereal going \o | 16:59 |
OXPHOS | wiking hey. I can't get the CMakeLists back..I squashed but the CMakeLists file has multiple changes | 17:00 |
OXPHOS | I think the feature branch needs rebase | 17:01 |
@wiking | mmm | 17:01 |
@wiking | wait | 17:01 |
@wiking | what exactly did you do :) | 17:01 |
@wiking | i mean from squash nothing else should happen | 17:02 |
@wiking | only that it becomes one simple commit | 17:02 |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 17:02 | |
-!- mode/#shogun [+o besser82] by ChanServ | 17:02 | |
OXPHOS | git reset --soft, then I noticed a line change (has several tabs/spaces before, but empty line now). So I want to change it back. | 17:03 |
OXPHOS | wiking: but it's hard to know how many spaces there're. So I grabbed the CMakeLists from shogun/develop | 17:03 |
OXPHOS | and replaced mine | 17:03 |
OXPHOS | wiking and Atom can automatically trim empty lines... | 17:19 |
-!- sanuj [~sanuj@117.204.255.134] has joined #shogun | 17:23 | |
sanuj | hello everyone! | 17:26 |
sanuj | got busy with some passport work today :/ | 17:27 |
arianepaola | hi sanuj | 17:27 |
sanuj | arianepaola, hi :) | 17:27 |
sanuj | i guess lisitsyn is not here | 17:27 |
OXPHOS | hey sanuj | 17:31 |
sanuj | OXPHOS, hi :) | 17:31 |
sanuj | what's up | 17:32 |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 258 seconds] | 17:33 | |
-!- leagoetz [~leagoetz@nat-247-179.internal.eduroam.ucl.ac.uk] has joined #shogun | 17:48 | |
sanuj | I'm going to write cookbook pages on LibSVR and neuralnets | 17:59 |
sanuj | arianepaola, OXPHOS Saurabh7 i hope there is no overlap ^^ | 17:59 |
arianepaola | sanuj: ok | 17:59 |
leagoetz | doesn't look like arianepaola: and Saurabh7: have done their cookbooks yet ... | 18:00 |
leagoetz | get going guys! | 18:00 |
leagoetz | it's an important deliveralbe | 18:00 |
leagoetz | able :) | 18:00 |
sanuj | hi lea :D | 18:00 |
sanuj | lisitsyn, there? | 18:00 |
leagoetz | heyhey | 18:01 |
lisitsyn | partially there | 18:02 |
lisitsyn | some lecture to start soon | 18:02 |
lisitsyn | sup? | 18:02 |
sanuj | okay | 18:03 |
sanuj | when will it get over? | 18:03 |
lisitsyn | 3 hours, got time right now | 18:03 |
lisitsyn | need any help? | 18:03 |
sanuj | yes | 18:03 |
sanuj | what is fire_updated? | 18:03 |
sanuj | do we keep typedef T type in shogun? | 18:03 |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 18:03 | |
-!- mode/#shogun [+o besser82] by ChanServ | 18:04 | |
OXPHOS | sanuj go ahead | 18:04 |
sanuj | OXPHOS, cool | 18:04 |
sanuj | lisitsyn, ^^ | 18:04 |
-!- leagoetz [~leagoetz@nat-247-179.internal.eduroam.ucl.ac.uk] has quit [] | 18:06 | |
sanuj | lisitsyn, i'll start a PR and start questions there | 18:07 |
sanuj | ask* | 18:07 |
arianepaola | wiking: this fixes it https://github.com/shogun-toolbox/shogun/pull/3212 | 18:12 |
@wiking | mmm i'm just wondeirng | 18:12 |
@wiking | what's the diff | 18:12 |
@wiking | between your ubuntu | 18:13 |
@wiking | and the buildbot's | 18:13 |
arianepaola | I am building with Python modular on | 18:14 |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 260 seconds] | 18:14 | |
arianepaola | cmake -GNinja -DENABLE_TESTING=ON -DPythonModular=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX="./install" .. | 18:15 |
arianepaola | ninja GoogleMock && ninja all && ninja install && ninja test | 18:15 |
@wiking | mmm | 18:21 |
arianepaola | wiking: a question regarding index_t | 18:26 |
arianepaola | it is defined in common.h as typedef int32_t index_t; | 18:26 |
arianepaola | wiking: is it used only for indexes >= 0 ? | 18:27 |
shogun-buildbot | build #6 of memleak - valgrind is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/memleak%20-%20valgrind/builds/6 | 18:42 |
-!- sanuj [~sanuj@117.204.255.134] has quit [Remote host closed the connection] | 19:47 | |
-!- glorb9 [d8338942@gateway/web/freenode/ip.216.51.137.66] has joined #shogun | 20:27 | |
glorb9 | Good afternoon. I'm having an odd cmake error that I've not seen before. | 20:27 |
glorb9 | Anyone here that might be able to give me a hand? | 20:27 |
glorb9 | Hello? | 20:30 |
-!- glorb9 [d8338942@gateway/web/freenode/ip.216.51.137.66] has quit [Quit: Page closed] | 20:32 | |
@wiking | arianepaola: yes | 21:19 |
@wiking | arianepaola: it should be unsigned | 21:19 |
@wiking | but then agian | 21:19 |
@wiking | there's this http://stackoverflow.com/questions/27275150/advice-on-unsigned-int-gangnam-style-edition | 21:20 |
@wiking | (google style) | 21:20 |
-!- lambday [8028b10a@gateway/web/freenode/ip.128.40.177.10] has quit [Quit: Page closed] | 21:24 | |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 21:58 | |
-!- mode/#shogun [+o besser82] by ChanServ | 21:58 | |
arianepaola | ok | 22:44 |
arianepaola | From https://google.github.io/styleguide/cppguide.html -> Equally bad bugs can occur when comparing signed and unsigned variables. | 22:49 |
arianepaola | wiking: it is probably best to change the comparison type in the tests to compare signed with signed variables | 22:50 |
lisitsyn | signed! | 22:55 |
lisitsyn | drop unsigned! | 22:58 |
arianepaola | yes, static_cast<int32_t>(strlen(const_char_array_variable)) | 23:16 |
arianepaola | bye everyone and have a great weekend | 23:40 |
--- Log closed Sat May 28 00:00:15 2016 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!