--- Log opened Wed Mar 22 00:00:14 2017 | ||
-!- norpi [~norbert.b@540391D6.catv.pool.telekom.hu] has quit [Read error: Connection reset by peer] | 00:07 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 00:12 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 00:51 | |
-!- goksinen_ [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 00:54 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Ping timeout: 258 seconds] | 00:57 | |
-!- goksinen_ [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 01:03 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 01:12 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 01:33 | |
-!- mikeling [uid89706@gateway/web/irccloud.com/x-rrtimeiwrndkxqdo] has joined #shogun | 01:52 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 02:09 | |
-!- Killin [~Killin-A1@2602:304:68a6:ddc0:1c90:463:a576:3640] has joined #shogun | 02:28 | |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has quit [Quit: Page closed] | 02:31 | |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has joined #shogun | 03:21 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 03:37 | |
-!- iRmantou [2d389488@gateway/web/freenode/ip.45.56.148.136] has joined #shogun | 04:01 | |
-!- OXPHOS_ [401e476d@gateway/web/freenode/ip.64.30.71.109] has joined #shogun | 04:32 | |
-!- neha [ca8e6c73@gateway/web/freenode/ip.202.142.108.115] has joined #shogun | 04:42 | |
neha | Hello! I'm currently working on a project (HPC's with unsupervised learning). I'm currently looking for a GSoc'17 internship project. I would like to start with this(https://github.com/shogun-toolbox/shogun/issues/2718) issue if it's still open! :) Can I? | 04:43 |
---|---|---|
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has quit [Ping timeout: 260 seconds] | 05:32 | |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has joined #shogun | 05:36 | |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has quit [Quit: Page closed] | 05:53 | |
@sukey | Pull Request #3736 "clean up memcpy in swig typemaps" merged by lambday - https://github.com/shogun-toolbox/shogun/pull/3736 | 06:23 |
@sukey | New Commit "Merge pull request #3736 from MikeLing/issuefix-3700 | 06:23 |
@sukey | clean up memcpy in swig typemaps" to shogun-toolbox/shogun by lambday: https://github.com/shogun-toolbox/shogun/commit/044780df172e4c48f71030254e94359beaa92d74 | 06:23 |
@sukey | Issue #3739 "add swig interface checks to appveyor" opened by vigsterkr - https://github.com/shogun-toolbox/shogun/issues/3739 | 06:36 |
@sukey | Issue #3739 "add swig interface checks to appveyor" vigsterkr added label: "SWIG" - https://github.com/shogun-toolbox/shogun/issues/3739 | 06:36 |
@sukey | Issue #3739 "add swig interface checks to appveyor" vigsterkr added label: "infrastructure" - https://github.com/shogun-toolbox/shogun/issues/3739 | 06:36 |
-!- neha [ca8e6c73@gateway/web/freenode/ip.202.142.108.115] has quit [Ping timeout: 260 seconds] | 06:37 | |
@sukey | Issue #3737 "CARTree can't be imported from modshogun" closed by vigsterkr - https://github.com/shogun-toolbox/shogun/issues/3737 | 06:38 |
-!- travis-ci [~travis-ci@ec2-54-147-54-125.compute-1.amazonaws.com] has joined #shogun | 07:19 | |
travis-ci | it's Soumyajit De'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/213706916 | 07:19 |
-!- travis-ci [~travis-ci@ec2-54-147-54-125.compute-1.amazonaws.com] has left #shogun [] | 07:19 | |
-!- OXPHOS_ [401e476d@gateway/web/freenode/ip.64.30.71.109] has quit [Ping timeout: 260 seconds] | 07:24 | |
shogun-buildbot | build #67 of nightly_default is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/67 blamelist: Viktor Gal <vigsterkr@gmail.com> | 07:45 |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has joined #shogun | 07:53 | |
iRmantou | Is there a problem with the subscription in mailing list? | 08:11 |
@wiking | mmm | 08:12 |
iRmantou | =。=! I can't subscribe it, It's just a blank page, Only I am in this case? | 08:17 |
@sukey | Issue #3737 "CARTree can't be imported from modshogun"- https://github.com/shogun-toolbox/shogun/issues/3737 | 08:24 |
-!- lambday|work [945742af@gateway/web/freenode/ip.148.87.66.175] has joined #shogun | 08:28 | |
-!- mode/#shogun [+o lambday|work] by ChanServ | 08:28 | |
mikeling | iRmantou: what's the meaning of a blank page? | 09:18 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has joined #shogun | 09:20 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 09:20 | |
iRmantou | I check the "Subscribe" under MAILING LIST on our website and there is only a page with | 09:33 |
iRmantou | no comtent | 09:33 |
iRmantou | content | 09:33 |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has quit [Ping timeout: 260 seconds] | 09:33 | |
@wiking | iRmantou, have you tried sending a mail to shogun-list-subscribe@shogun-toolbox.org | 09:34 |
@wiking | ? | 09:34 |
iRmantou | no, I just want to subscribe to the mailing list becauseI see a requirement. | 09:39 |
iRmantou | https://summerofcode.withgoogle.com/organizations/6322302468751360/#work-with-us | 09:39 |
iRmantou | "Subscribe to the mailing list and introduce yourself" | 09:39 |
iRmantou | I have tested other organizations' mailing list, they are ok, so I think this is may not my net problems. | 09:41 |
@wiking | iRmantou, just send an email to the address i've given you | 09:46 |
@wiking | that should get you subscribed | 09:46 |
iRmantou | ok ! thank you! | 09:46 |
@wiking | CaBa, ping | 09:50 |
@wiking | wtf just happened here what do you think | 09:50 |
@wiking | make[2]: *** No rule to make target 'GoogleMock/sbc/GoogleMock/googletest/include/gtest/gtest_prod.h', needed by 'tests/unit/CMakeFiles/shogun-unit-test.dir/io/LibSVMFile_unittest.cc.o'. Stop. | 09:50 |
@wiking | :))) | 09:50 |
mikeling | wiking: How could I register an instance of my global fixture? call AddGlobalTestEnvironment in main() of discover_gtest_tests.cpp? | 09:55 |
@wiking | why do you want to add it to the discover gtests? | 09:56 |
@wiking | that cpp is just generating the ctest stuff | 09:56 |
mikeling | I just founding the entry of all tests, I mean something like RUN_ALL_TESTS | 09:57 |
@wiking | ? | 09:59 |
@wiking | you need to add the global fixture to shogun-unit-test | 09:59 |
@wiking | executable | 09:59 |
mikeling | by cmakelists? | 10:00 |
@HeikoS | wiking: jojo | 10:00 |
@HeikoS | wiking: I really wonder, how the conference was | 10:01 |
@wiking | what? | 10:01 |
@wiking | ah | 10:01 |
@wiking | not many people | 10:01 |
@wiking | but was ok | 10:01 |
@wiking | :) | 10:01 |
@HeikoS | how was the talk=? | 10:01 |
@wiking | sorry i didn't answer till now | 10:01 |
@wiking | because i was busy with other stuff | 10:01 |
@HeikoS | nw | 10:01 |
@wiking | in what way? | 10:01 |
@wiking | i spoke a lot :) | 10:01 |
@wiking | run out of time :) | 10:01 |
@wiking | was casual | 10:01 |
@HeikoS | feedback? | 10:02 |
@wiking | mmm 2 possible gsoc applicants | 10:02 |
@wiking | the session leader like the talk | 10:02 |
@wiking | as well as another guy | 10:02 |
@wiking | but again | 10:02 |
@wiking | the talk was at midday | 10:02 |
@wiking | :DD | 10:02 |
@wiking | lunch time :P | 10:02 |
@HeikoS | hehe i see | 10:02 |
iRmantou | wait me! I am writing the applicant! | 10:09 |
@HeikoS | wiking: btw | 10:09 |
@wiking | eys | 10:09 |
@HeikoS | wiking: I will be in Tokyo from Friday to end of month | 10:09 |
@wiking | *yes | 10:09 |
@wiking | ah yeah you told me | 10:09 |
@HeikoS | I hope to have time to feedback the applications | 10:09 |
@HeikoS | its a bad clash | 10:09 |
@wiking | ah well 2 things | 10:10 |
@HeikoS | will try to do some stuff today | 10:10 |
@wiking | one is about the org shits | 10:10 |
@wiking | you know that we got accepted to : ) | 10:10 |
@HeikoS | wiking: we are waiting for the FSA | 10:10 |
@wiking | yeah i guess | 10:10 |
@HeikoS | Ill ping again | 10:10 |
@wiking | we'll get it when you are in tokyo | 10:10 |
@wiking | :) | 10:10 |
@HeikoS | yeah we need to sign it all | 10:10 |
@HeikoS | so not sure how that will work, digitally or mail | 10:10 |
@wiking | HeikoS, i reckon they're gonna send it after the 2nd board meeting | 10:10 |
@HeikoS | yep something like that | 10:10 |
@wiking | that supposed to be on the 23rd | 10:11 |
@wiking | or somethin | 10:11 |
@wiking | anyhow yea | 10:11 |
@HeikoS | wiking: also wanted to ask you to feedback the PR in doc by lea | 10:11 |
@HeikoS | with "mission" | 10:11 |
@wiking | let's see how we'll have to circulate stuff | 10:11 |
@wiking | ah | 10:11 |
@wiking | will check on it tomo | 10:11 |
@HeikoS | whether you agree, have more things to add | 10:11 |
@HeikoS | no need for detailed comments, more like big picture | 10:11 |
@HeikoS | would be cool to get that out before we get exposed via the org | 10:11 |
@HeikoS | wiking: we really need to discuss this tags business | 10:12 |
@HeikoS | we also need transient | 10:12 |
@HeikoS | and we need to discuss where the members are stored | 10:13 |
@HeikoS | CaBa: jo | 10:13 |
@HeikoS | https://github.com/shogun-toolbox/shogun/pull/3721/files#diff-9c3599c0d2090e493be261b079e9b63eR715 | 10:13 |
@wiking | HeikoS, the shitty part of *transient that there's no support for such thing in c++ :) | 10:15 |
@wiking | i mean the nearest thing would be attributes | 10:16 |
@wiking | http://en.cppreference.com/w/cpp/language/attributes | 10:16 |
@wiking | but those are not customizable by far :( | 10:16 |
@HeikoS | wiking: mmh I guess tags can do that | 10:16 |
@HeikoS | it is just parameters that are get/settable, but wont be serialized and wont be part of equals | 10:16 |
@wiking | ah you mean to give a special wrapper | 10:16 |
@wiking | Transient<Type> | 10:16 |
@wiking | and add that to the tag? | 10:16 |
@HeikoS | yep | 10:16 |
@wiking | ah that's possible | 10:16 |
@wiking | but then again | 10:16 |
@HeikoS | yeah it is | 10:17 |
@HeikoS | it is easy | 10:17 |
@wiking | the story is still that | 10:17 |
@HeikoS | just need a flag | 10:17 |
@HeikoS | (and it is independent of where they are stored) | 10:17 |
@wiking | wtf we do when people dont actually | 10:17 |
@wiking | register stuff :D | 10:17 |
lisitsyn | it fails | 10:17 |
@HeikoS | get/set fails | 10:17 |
@wiking | lisitsyn, because you presume | 10:17 |
@wiking | that everybody is gonna use | 10:17 |
@wiking | get/set | 10:17 |
lisitsyn | yes | 10:17 |
lisitsyn | you have no other way | 10:17 |
lisitsyn | :D | 10:17 |
lisitsyn | resistance is futile | 10:17 |
@wiking | hahahha | 10:17 |
@HeikoS | if you dont make the vars members then we have it | 10:18 |
@HeikoS | Buutttt | 10:18 |
@HeikoS | hackathon stuff then | 10:18 |
lisitsyn | using only get/set open a whole fucking window of opportuinites | 10:18 |
lisitsyn | like observing what happens while you train | 10:18 |
@wiking | :) | 10:19 |
lisitsyn | ah and we also can have a tensorboard-like thing | 10:19 |
@wiking | lisitsyn, how you hook anything to an unordered_map? | 10:19 |
lisitsyn | hook what? | 10:19 |
@wiking | i mean then the get/set should be annotatable :) | 10:20 |
@wiking | like aspectj | 10:20 |
@wiking | :P | 10:20 |
lisitsyn | do not get it yet | 10:20 |
@wiking | well look | 10:20 |
@wiking | you are saying | 10:20 |
@HeikoS | wiking: you mean for type safety? :D | 10:20 |
@wiking | like observing what happens while you train | 10:20 |
lisitsyn | yes | 10:20 |
@wiking | "like observing what happens while you train" | 10:20 |
@wiking | how do you hook anything | 10:20 |
@wiking | so you can monitor | 10:20 |
@sukey | Issue #3739 "add swig interface checks to appveyor" karlnapf added label: "entrance" - https://github.com/shogun-toolbox/shogun/issues/3739 | 10:20 |
lisitsyn | map<Tag, Any> | 10:20 |
lisitsyn | Any can tell its json | 10:20 |
@wiking | que? | 10:21 |
lisitsyn | so you get a json of parameters | 10:21 |
@wiking | but still | 10:21 |
@wiking | how do you monitor this runtime? | 10:21 |
lisitsyn | {"parameter": <something generated by any>} | 10:21 |
lisitsyn | you just register all objects | 10:21 |
lisitsyn | somewhere | 10:21 |
lisitsyn | and then iterate over all objects over all parameters | 10:21 |
@wiking | ok | 10:21 |
@wiking | i think we are | 10:21 |
lisitsyn | should be possible | 10:21 |
@wiking | totally talking to another | 10:21 |
@wiking | universe :) | 10:22 |
@wiking | man | 10:22 |
@wiking | you are saying | 10:22 |
@wiking | that | 10:22 |
@wiking | "like observing what happens while you train" = for me this means that you can actually somehow hook to ALL the parameters of an object and stream their values (when they change) | 10:22 |
lisitsyn | no I mean retrieving the | 10:22 |
@wiking | and i guess they change when you | 10:22 |
lisitsyn | you just retrieve the current state | 10:23 |
@wiking | insert into the map | 10:23 |
@wiking | lol but man | 10:23 |
@wiking | how's that monitoring? | 10:23 |
lisitsyn | observer is much more difficult | 10:23 |
@wiking | :D | 10:23 |
@wiking | unordered_map is not threadsafe | 10:23 |
@wiking | so you just get like | 10:23 |
@wiking | super weird | 10:23 |
@wiking | snapshots | 10:23 |
@wiking | right? | 10:23 |
@wiking | i mean that's not really "observing what happens while you train" | 10:23 |
lisitsyn | well we can use thread safe map :P | 10:24 |
@wiking | but even then | 10:24 |
@wiking | you gonna be iterating | 10:24 |
lisitsyn | it's not the detail ruining the whole stuff | 10:24 |
@wiking | like while(1) ? | 10:24 |
@wiking | :D | 10:24 |
lisitsyn | yes if you want to | 10:24 |
@wiking | no i dont want to | 10:24 |
@wiking | i'm just asking questions | 10:24 |
@wiking | because your statement is that | 10:24 |
@wiking | "observing what happens while you train" | 10:24 |
@wiking | :) | 10:24 |
lisitsyn | yes and what's wrong? | 10:24 |
@wiking | and i've just showed you | 10:25 |
@wiking | that it's not really the case | 10:25 |
@wiking | if one would like to do that | 10:25 |
@wiking | the best would be to have something at setter part | 10:25 |
lisitsyn | yes you can observe snapshots but I am ok with snapshots | 10:25 |
@wiking | lisitsyn, and i presume the presumption in this case is | 10:25 |
@wiking | that you cannot store references in the tags | 10:25 |
@wiking | right? | 10:25 |
lisitsyn | why so? | 10:26 |
@wiking | well because | 10:26 |
@wiking | then it's even more so | 10:26 |
@wiking | :) | 10:26 |
@wiking | that you have no idea | 10:26 |
@wiking | what's happening with the parameters | 10:26 |
@wiking | :> | 10:26 |
@wiking | as they are mutable | 10:26 |
lisitsyn | gosh man we can just implement an observer if you're so obsessed :D | 10:26 |
@wiking | no i'm not obsessed | 10:27 |
@wiking | but please | 10:27 |
@wiking | be a bit more precise | 10:27 |
@wiking | when you have a statement about | 10:27 |
@wiking | what tags can do | 10:27 |
lisitsyn | yes tags still can do it | 10:27 |
@wiking | i see :) | 10:28 |
@wiking | cool | 10:28 |
lisitsyn | if you register things you can track them right? | 10:28 |
lisitsyn | that's why tensorboard works | 10:28 |
@wiking | true | 10:28 |
@wiking | but that's true atm as well | 10:28 |
@wiking | right? | 10:28 |
@wiking | i mean we do register shit atm as well | 10:29 |
@wiking | (see SG_ADD) | 10:29 |
lisitsyn | yes but you can't track set now | 10:29 |
lisitsyn | with tags you can | 10:29 |
@wiking | but the point was above | 10:29 |
@wiking | that you dont actually have to do set | 10:29 |
@wiking | with tags either | 10:29 |
@wiking | right? | 10:29 |
@wiking | just as in case of SG_ADD | 10:29 |
@sukey | Issue #3740 "Create data generator class for unit testing" opened by karlnapf - https://github.com/shogun-toolbox/shogun/issues/3740 | 10:30 |
@sukey | Issue #3740 "Create data generator class for unit testing" karlnapf added label: "Cleanups" - https://github.com/shogun-toolbox/shogun/issues/3740 | 10:30 |
@sukey | Issue #3740 "Create data generator class for unit testing" karlnapf added label: "testing" - https://github.com/shogun-toolbox/shogun/issues/3740 | 10:30 |
@sukey | Issue #3740 "Create data generator class for unit testing" karlnapf added label: "entrance" - https://github.com/shogun-toolbox/shogun/issues/3740 | 10:30 |
-!- iRmantou_ [55cb2405@gateway/web/freenode/ip.85.203.36.5] has joined #shogun | 10:30 | |
lisitsyn | why don't you? | 10:30 |
@wiking | ? | 10:31 |
lisitsyn | why don't you have to do set? what does it mean? | 10:31 |
@wiking | well | 10:31 |
@wiking | if i store a reference | 10:31 |
@wiking | or a pointer | 10:31 |
@wiking | i dont know what's happening anymore with the value per se | 10:31 |
@wiking | but this is true for the current case as well | 10:31 |
lisitsyn | oh I don't like storing a reference | 10:31 |
@wiking | lisitsyn, yeah but i've just asked | 10:31 |
lisitsyn | I think we're talking about reference to other object | 10:31 |
@wiking | 20 lines above :) | 10:31 |
lisitsyn | storing a pointer is no go | 10:31 |
@wiking | and then? | 10:32 |
@wiking | if i have to? | 10:32 |
@wiking | Transient<> ?: ) | 10:32 |
lisitsyn | nah transient is other thing | 10:32 |
@wiking | i know | 10:32 |
@wiking | but again | 10:32 |
@wiking | what do i do? | 10:32 |
lisitsyn | storing pointer = statefullness that's not tracked | 10:33 |
@wiking | if i want to/need to | 10:33 |
@wiking | because | 10:33 |
@wiking | <lisitsyn> oh I don't like storing a reference | 10:33 |
@wiking | <lisitsyn> storing a pointer is no go | 10:33 |
@wiking | might be valid | 10:33 |
@wiking | but still | 10:33 |
@wiking | there could be cases | 10:33 |
@wiking | where there's no other option right? | 10:33 |
lisitsyn | no, why would you need that? | 10:33 |
@wiking | :) | 10:33 |
@wiking | so you are saying | 10:34 |
@wiking | that no matter what | 10:34 |
@wiking | nobody wants or shall store a point | 10:34 |
@wiking | as an attribute of a class? | 10:34 |
@wiking | *pointer | 10:34 |
lisitsyn | sure, why to store a pointer? it's broken from the beginning | 10:34 |
@wiking | :) | 10:34 |
@wiking | this is too good man | 10:34 |
@wiking | :) | 10:34 |
-!- iRmantou [2d389488@gateway/web/freenode/ip.45.56.148.136] has quit [Ping timeout: 260 seconds] | 10:34 | |
lisitsyn | you need to store pointers in sgvector and stuff like that | 10:34 |
lisitsyn | but in 'user' code? | 10:34 |
@wiking | http://www.quickmeme.com/img/9f/9f97f7502ecc5767b455e9ecc3978da0fe9bb10d351931c98569e7af8fd82920.jpg | 10:35 |
@wiking | i couldn't resist sorry | 10:35 |
@wiking | :) | 10:35 |
lisitsyn | well I'd love to see a kernel that needs to store a pointer | 10:35 |
lisitsyn | or a classifier | 10:35 |
@HeikoS | maybe we can have fistfights in zurich | 10:36 |
lisitsyn | we're talking about shogun classes right? | 10:37 |
@wiking | ? what fight | 10:37 |
lisitsyn | I am not into that pervert things | 10:37 |
lisitsyn | :D | 10:37 |
@wiking | i guess we are discussing here | 10:37 |
lisitsyn | wiking: just one example why would you need to store a pointer in a *shogun class*? | 10:37 |
@wiking | lisitsyn, i mean i just find your comment of | 10:37 |
@wiking | not storing pointer | 10:37 |
@wiking | as strong | 10:38 |
@wiking | as 64k should be enough for everybody | 10:38 |
lisitsyn | yes you have to limit something sometimes but we're talking about some classifiers and stuff | 10:38 |
@wiking | yeah sure you can argue as much as you want about | 10:38 |
@wiking | why it's stupid | 10:38 |
@wiking | but still | 10:38 |
@wiking | that's one of the simplest features | 10:38 |
@wiking | of the language | 10:38 |
lisitsyn | it's never used in shogun now | 10:38 |
@wiking | now you just wanna ignore that whole part | 10:38 |
@wiking | sure do it | 10:38 |
@wiking | but i'm not convinved | 10:39 |
@wiking | *convinced that it's the smartest move to do | 10:39 |
lisitsyn | well you may allow tags to pointers if you want to but I have no use case for that | 10:39 |
@wiking | HeikoS, "A unit test for single a multi threads of a kernel machine would help | 10:40 |
@wiking | " | 10:40 |
@wiking | hahahaha | 10:40 |
@wiking | too good | 10:40 |
@wiking | :> | 10:40 |
lisitsyn | wiking: we also have to limit what classes can we store | 10:40 |
@wiking | too bad that half of our CIs dont do well with multithreads :) | 10:40 |
lisitsyn | because they are all exported to swig | 10:41 |
@HeikoS | should actually test all machines for single and multithread train | 10:41 |
@HeikoS | at least on buildbot | 10:41 |
@wiking | HeikoS, write the code :) | 10:41 |
@HeikoS | wiking: yeah know | 10:41 |
@wiking | HeikoS, they'll do it for you :) | 10:41 |
@HeikoS | wiking: just a bit nervous about these patches, they are hard to parse | 10:41 |
@wiking | HeikoS, which? | 10:42 |
@HeikoS | wiking: openmp port of pthread | 10:42 |
@wiking | HeikoS, mmm man | 10:42 |
@wiking | :))) | 10:42 |
@wiking | let me tell you a secret | 10:42 |
@wiking | :D | 10:42 |
lisitsyn | can we start using std::thread instead? | 10:42 |
@wiking | those are copy-paste solutions from micmn | 10:42 |
@wiking | :> | 10:42 |
@HeikoS | i know | 10:42 |
@HeikoS | even worse | 10:42 |
@wiking | i'm not taking those as real contributions | 10:42 |
@wiking | :D | 10:42 |
@HeikoS | lisitsyn: hackathon :D | 10:42 |
@wiking | (i hope everybody hears this :P) | 10:42 |
lisitsyn | what does mr. wiking think about std::thread? | 10:42 |
@wiking | lisitsyn, ? sorry Sergey did i offendend you by any way? | 10:43 |
@wiking | *offended | 10:43 |
@wiking | or i dont get really the sarcasm :) | 10:43 |
lisitsyn | just curious | 10:43 |
@wiking | other than that i believe i had the first patch to shogun using std::thread when i added windozer support | 10:43 |
@wiking | evne asked u to review | 10:43 |
lisitsyn | I have no memory | 10:44 |
@wiking | that's ok | 10:44 |
lisitsyn | like lstm but lst | 10:44 |
@wiking | but yeah other than input parsers | 10:44 |
@wiking | or such | 10:44 |
@wiking | i would try to use openmp | 10:44 |
@wiking | if possible | 10:44 |
lisitsyn | I just mean openmp is not that good sometimes | 10:44 |
lisitsyn | because it is highly implicit | 10:44 |
@wiking | yeah but most of the times is sufficient | 10:44 |
@wiking | shit like that has been doen | 10:44 |
@wiking | with pthreads | 10:44 |
@wiking | for kernel machines | 10:44 |
@wiking | or kernel | 10:44 |
@wiking | openmp is more than sufficient | 10:45 |
@wiking | and really lowers the LoC | 10:45 |
lisitsyn | well at least you're explicit what happens | 10:45 |
lisitsyn | :D | 10:45 |
@wiking | lisitsyn, yeah but see the LoC? | 10:45 |
@wiking | :)))) | 10:45 |
lisitsyn | because I have no idea about # threads and stuff like that | 10:45 |
@HeikoS | i think xvalidation might be nice to do with something more sophisticated | 10:45 |
@wiking | HeikoS, heheh xval shoudl be thrown out | 10:45 |
@wiking | in it's current form | 10:45 |
@wiking | :) | 10:45 |
@wiking | that's a hack a patch | 10:45 |
@wiking | and i know tha tyou know | 10:45 |
@HeikoS | just revert revert revert it | 10:45 |
@wiking | :D | 10:45 |
@HeikoS | then we can release | 10:45 |
@HeikoS | :D | 10:45 |
@wiking | because we were fucking with that | 10:45 |
@wiking | since october | 10:46 |
@HeikoS | yeah lets kill this guy | 10:46 |
@HeikoS | and then do in a feature branch with some investigation | 10:46 |
@HeikoS | like testing it on all classes etc | 10:46 |
@HeikoS | the stuff CaBa found out | 10:46 |
lisitsyn | I need some help with tags | 10:46 |
@HeikoS | guy=patch | 10:46 |
@wiking | :>>>>>>>>>>>>>>>>>>>> | 10:46 |
@HeikoS | :D | 10:46 |
@wiking | i'm fine with both | 10:46 |
@wiking | :D | 10:46 |
lisitsyn | help me please :P | 10:47 |
@wiking | lisitsyn, u need xval solution? :) | 10:47 |
lisitsyn | I need to compile shogun | 10:47 |
@wiking | make | 10:47 |
@wiking | or use ninja | 10:47 |
@wiking | better :) | 10:47 |
lisitsyn | each time I try | 10:47 |
@wiking | really faster | 10:47 |
lisitsyn | well I mean I update my git | 10:47 |
lisitsyn | and each time I get a new error | 10:47 |
lisitsyn | and while I'm investigating | 10:47 |
@wiking | mmm | 10:47 |
@wiking | it should compile | 10:48 |
@wiking | on osx | 10:48 |
@wiking | i reckon that's what you use | 10:48 |
@wiking | so u dont even have the problem | 10:48 |
@wiking | of multithreading | 10:48 |
@wiking | :DDDDDDDDDDDDDDDDDDDDDDDDDDDDDD | 10:48 |
@wiking | (openmp just to be specific) | 10:48 |
lisitsyn | yeah I have no problem with shogun in general | 10:48 |
lisitsyn | because I can't compile it | 10:48 |
lisitsyn | most of the times | 10:48 |
lisitsyn | :D | 10:48 |
@wiking | good | 10:48 |
@wiking | r u using develop? :) | 10:48 |
lisitsyn | well yes | 10:48 |
@wiking | or some super branch like HeikoS last time ? ) | 10:49 |
@wiking | ok so what's your error? | 10:49 |
lisitsyn | well its different each time :D | 10:49 |
@wiking | ok | 10:49 |
lisitsyn | let me check what is it now | 10:49 |
@wiking | for me | 10:49 |
@wiking | it's enough | 10:49 |
@wiking | what is it now | 10:49 |
@wiking | :) | 10:49 |
@wiking | meaning i dong give a damn about your whining how everything is a shit.... i know! :) | 10:49 |
lisitsyn | no no I am not whining | 10:50 |
lisitsyn | its me who can fix it as well | 10:50 |
@wiking | so what's the erorr? | 10:50 |
@HeikoS | I had so many problem compiling things :D | 10:50 |
@HeikoS | config.h is fucked | 10:50 |
@wiking | HeikoS, not at all | 10:50 |
@wiking | that thing is finally fixed | 10:50 |
@wiking | :D | 10:50 |
@HeikoS | had to manually edit | 10:50 |
@HeikoS | when? | 10:50 |
-!- micmn [uid216702@gateway/web/irccloud.com/x-jwnhkzyvxkgkmtsm] has joined #shogun | 10:50 | |
@wiking | HeikoS, buuuuuuuuuuuuuut! | 10:50 |
@HeikoS | I had a lot of issues changing options last week | 10:51 |
@HeikoS | but doesnt matter | 10:51 |
@wiking | do you have | 10:51 |
@wiking | src/shogun/lib/config.h ? | 10:51 |
@HeikoS | it is like when you change options | 10:51 |
@HeikoS | it doesnt update the config in build dir | 10:51 |
@HeikoS | or vice versa, I forgot | 10:51 |
@wiking | which option is it? | 10:51 |
@HeikoS | it was all those | 10:51 |
@wiking | becuase i beleive | 10:51 |
@HeikoS | JSON, SMVLIGHT | 10:51 |
@HeikoS | etc | 10:51 |
@wiking | that should not be the case | 10:51 |
@wiking | :) | 10:51 |
@wiking | configure_file works | 10:51 |
@wiking | :> | 10:51 |
lisitsyn | wiking: oh ok | 10:51 |
lisitsyn | error: unknown type name 'CWrappedSGMatrix' | 10:51 |
lisitsyn | case PT_INT8: return new CWrappedSGMatrix<int8_t>(); | 10:51 |
@HeikoS | I can reproduce at somepoint | 10:51 |
lisitsyn | that's what I get now | 10:51 |
@wiking | ah that's each | 10:51 |
@HeikoS | it is changing | 10:51 |
@wiking | *easy | 10:51 |
@HeikoS | yeah that is one | 10:51 |
@HeikoS | lisitsyn: that has to do with config | 10:52 |
@wiking | lisitsyn, ls -la ./src/shogun/lib/config.h | 10:52 |
@HeikoS | because the class is guarded when testing is disabled | 10:52 |
@wiking | what's the output | 10:52 |
@wiking | (not in build but in the root of the cloned repo) | 10:52 |
lisitsyn | ls -la ./src/shogun/lib/config.h [12:50:29] | 10:52 |
lisitsyn | -rw-r--r-- 1 lisitsyn LD\Domain Users 3938 Feb 15 15:55 ./src/shogun/lib/config.h | 10:52 |
lisitsyn | ehm? :) | 10:52 |
@wiking | domain users :) | 10:53 |
@wiking | nice | 10:53 |
@wiking | :> | 10:53 |
@wiking | ok | 10:53 |
@sukey | Issue #3000 "Replace an undocumented example by a cookbook page" closed by karlnapf - https://github.com/shogun-toolbox/shogun/issues/3000 | 10:53 |
@wiking | so this is in the root of the src | 10:53 |
@wiking | right? | 10:53 |
@wiking | not in build dir | 10:53 |
lisitsyn | not in build | 10:53 |
@wiking | if so then do me a favour | 10:53 |
lisitsyn | I enabled testing | 10:53 |
@wiking | rm ./src/shogun/lib/config.h | 10:53 |
lisitsyn | git is clean | 10:53 |
@wiking | because that file should not be there anymore | 10:53 |
lisitsyn | ok then I should rebase that branch | 10:54 |
@wiking | nono | 10:54 |
@wiking | it's fine | 10:54 |
@wiking | i mean what's your | 10:54 |
@wiking | HEAD (before your shit) | 10:54 |
@wiking | sha1 | 10:54 |
@wiking | :) | 10:54 |
@wiking | anyhow the problem is that | 10:54 |
@wiking | src/shogun/lib/config.h is in .gitignore | 10:54 |
@wiking | i reckon we should remove it now from there | 10:54 |
lisitsyn | I am on feature/private-wector | 10:54 |
@wiking | that's published? | 10:54 |
@wiking | right | 10:54 |
lisitsyn | yes | 10:54 |
lisitsyn | it's a PR | 10:55 |
@wiking | pr? or feature | 10:55 |
@wiking | sergey | 10:55 |
@wiking | those are xor | 10:55 |
@wiking | remember the conversation | 10:55 |
@wiking | :)))) | 10:55 |
@wiking | PLEASE DO NOT PR a feature branch | 10:55 |
@wiking | :) | 10:55 |
lisitsyn | why? | 10:55 |
@wiking | because it'll make CI double work | 10:55 |
lisitsyn | ah yeah I remember | 10:55 |
@HeikoS | gotta go now | 10:55 |
@wiking | :) | 10:55 |
@wiking | cool | 10:55 |
@HeikoS | see you later lovers | 10:55 |
lisitsyn | but | 10:55 |
lisitsyn | how | 10:55 |
lisitsyn | can | 10:55 |
lisitsyn | I | 10:55 |
lisitsyn | :D | 10:55 |
@wiking | :> | 10:55 |
@wiking | remember? | 10:55 |
@wiking | :) | 10:55 |
lisitsyn | wiking: you mean commit to my fork? | 10:56 |
@wiking | yes | 10:56 |
lisitsyn | is it what I should do? | 10:56 |
lisitsyn | ohk next time | 10:56 |
@wiking | no not the fork | 10:56 |
@wiking | i mean is this a feature branch of shogun | 10:56 |
@wiking | ? | 10:56 |
@wiking | or on yours | 10:56 |
lisitsyn | uhmm yes? | 10:56 |
@wiking | ah if shogun then it's all good | 10:56 |
@wiking | it'll be picked up by all the CI | 10:56 |
@wiking | so anyhow | 10:56 |
lisitsyn | https://github.com/shogun-toolbox/shogun/tree/feature/private-wector | 10:56 |
lisitsyn | that's the branch | 10:57 |
lisitsyn | I want to test it | 10:57 |
@wiking | okok | 10:57 |
@wiking | so no need to rebase | 10:57 |
@wiking | just fucking | 10:57 |
lisitsyn | but I fail miserably | 10:57 |
@wiking | rm ./src/shogun/lib/config.h | 10:57 |
lisitsyn | becoz I am stoopid | 10:57 |
lisitsyn | I removed it | 10:57 |
@wiking | ok | 10:57 |
@wiking | now just for the sake | 10:57 |
lisitsyn | make[2]: *** No rule to make target `../src/shogun/lib/config.h', needed by `src/shogun/base/class_list.cpp'. Stop. | 10:57 |
lisitsyn | now I can't even make | 10:57 |
lisitsyn | cries deeply | 10:57 |
lisitsyn | what have you done to me | 10:57 |
@wiking | ok but man wait | 10:57 |
@wiking | in build dir | 10:57 |
@wiking | ls ./src/shogun/lib/config.h | 10:58 |
@wiking | lisitsyn, fuck can u start a tmate | 10:58 |
@wiking | and i can do this for you | 10:58 |
@wiking | it's gonna be much faster | 10:58 |
@wiking | brew install tmate | 10:58 |
lisitsyn | what's tmate | 10:58 |
@wiking | and send me in private the ssh | 10:58 |
@wiking | lisitsyn, https://tmate.io/ | 10:58 |
@wiking | welcome to 2017 | 10:58 |
@wiking | :> | 10:58 |
lisitsyn | not sure if my employer would like that | 10:58 |
lisitsyn | :D | 10:58 |
@wiking | ah ok | 10:58 |
@wiking | :) | 10:58 |
@wiking | then we gonna be doing this for the next hour | 10:59 |
@wiking | :)))) | 10:59 |
@wiking | but yeha ok so | 10:59 |
lisitsyn | wiking: well I removed and cmake again | 10:59 |
@wiking | ls build/src/shogun/lib/config.h | 10:59 |
lisitsyn | now it tries to compile | 10:59 |
@wiking | okok | 10:59 |
@wiking | so the thing is that those shit happened | 10:59 |
@wiking | when the config.h was getting out of state | 10:59 |
lisitsyn | so what happened to me? | 10:59 |
@wiking | because i moved config.h generation to build | 10:59 |
@wiking | but because of that | 10:59 |
-!- HeikoS [~heiko@host-92-0-178-129.as43234.net] has quit [Ping timeout: 240 seconds] | 10:59 | |
@wiking | everybody who has been using shogun | 10:59 |
lisitsyn | ok so it clashes | 11:00 |
@wiking | for longer time | 11:00 |
@wiking | would have some shit in src/shogun/lib/ | 11:00 |
@wiking | and because gitignore has config.h | 11:00 |
@wiking | if you do | 11:00 |
@wiking | git status | 11:00 |
@wiking | you think you are in sync | 11:00 |
@wiking | :) | 11:00 |
@wiking | these lines should go from gitignore | 11:01 |
@wiking | /src/shogun/lib/config.h | 11:01 |
lisitsyn | don't we remove it from gitignore | 11:01 |
lisitsyn | because it is like | 11:01 |
@wiking | ah fuck | 11:01 |
@wiking | yeah | 11:01 |
@wiking | man | 11:01 |
@wiking | one more thing | 11:01 |
@wiking | src/shogun/base/class_list.cpp | 11:01 |
@wiking | remove this as well | 11:01 |
@wiking | :) | 11:01 |
@wiking | i reckon you have such thing as well | 11:01 |
@wiking | right? | 11:01 |
@wiking | (not in build but actual src | 11:02 |
lisitsyn | oh | 11:02 |
@wiking | see what i mean? :) | 11:02 |
@wiking | can i remove those from ignore? | 11:02 |
lisitsyn | we should remove it from ignore for sure | 11:02 |
lisitsyn | it's really confusing | 11:02 |
@wiking | yep | 11:02 |
@wiking | i mean it was a shitty solution | 11:03 |
@wiking | from the beginning | 11:03 |
lisitsyn | yes one should not touch src dir | 11:03 |
@wiking | but now | 11:03 |
@wiking | it's really out of src build | 11:03 |
@wiking | so you can have the build dir wherevers | 11:03 |
lisitsyn | wiking: versionstring? | 11:04 |
lisitsyn | is it out as well? | 11:04 |
@wiking | mmm | 11:04 |
@wiking | good question | 11:04 |
@wiking | i would hope so | 11:04 |
@wiking | dont remember the patch tough | 11:04 |
@wiking | yes it is | 11:04 |
@wiking | ok | 11:04 |
@wiking | i'm sending a commit | 11:04 |
@sukey | New Commit "dont ignore class_list.cpp versionstring.h and config.h" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/10c29e48f511dd1dc606eec9ea83eaaf014a3c2d | 11:05 |
@wiking | there | 11:05 |
@wiking | :) | 11:05 |
lisitsyn | oh you can be so kind | 11:06 |
@wiking | so i reckon | 11:06 |
@wiking | if you clearup those remainers | 11:06 |
@wiking | *remainders | 11:06 |
@wiking | things should get into a better shape | 11:06 |
@sukey | Pull Request #3735 "Move generated serialization unit tests into separate executable" merged by vigsterkr - https://github.com/shogun-toolbox/shogun/pull/3735 | 11:06 |
@sukey | New Commit "Merge pull request #3735 from micmn/feature/serialization-binary | 11:06 |
@sukey | Move generated serialization unit tests into separate executable" to shogun-toolbox/shogun by vigsterkr: https://github.com/shogun-toolbox/shogun/commit/4584965a2fa85d3a576662e7221efa406337a1d7 | 11:06 |
@wiking | lisitsyn, i gotta go now.... but i'll be around in an hour or so | 11:07 |
@wiking | if you still have compilation troubles lemm eknow | 11:07 |
@wiking | btw | 11:07 |
lisitsyn | wiking: ok thanks | 11:07 |
lisitsyn | well I usually try to resolve them by myself | 11:07 |
@wiking | i would suggest you to use a server to compile shogun | 11:07 |
lisitsyn | but then I am sucked into problems | 11:07 |
@wiking | it's faster better :) | 11:07 |
lisitsyn | wiking: well | 11:07 |
lisitsyn | wiking: it is true | 11:07 |
@wiking | lisitsyn, you have account on both fatbot | 11:07 |
@wiking | and i believe on hedonism as well | 11:07 |
lisitsyn | wiking: yeah yeah not an issue I have a lot of machines | 11:08 |
lisitsyn | there is different thing | 11:08 |
@wiking | drwxr-xr-x 5 lisitsyn lisitsyn 4096 Sep 10 2015 lisitsyn | 11:08 |
lisitsyn | it'd be cool to have it compiling on mac | 11:08 |
@wiking | ah yeah | 11:08 |
@wiking | sure it does | 11:08 |
* wiking compiles on mac | 11:08 | |
@wiking | so it works most def | 11:08 |
lisitsyn | but it would be faster to compile on some ubuntu server for sure | 11:08 |
@wiking | some of the unit tests might fail | 11:08 |
@wiking | but yeah i use servers mostly | 11:09 |
lisitsyn | wiking: that's why I was thinking about dist build | 11:09 |
@wiking | lisitsyn, hahah bitch | 11:09 |
@wiking | i will not devops you that | 11:09 |
@wiking | :)))) | 11:09 |
lisitsyn | sure you don't | 11:09 |
@wiking | i'm tired man of shogun devops | 11:09 |
lisitsyn | never asked you :) | 11:09 |
@wiking | hahah | 11:09 |
@wiking | i remember the fuckin gday | 11:09 |
@wiking | when you just wriote | 11:09 |
lisitsyn | I just ask if you think it is a good idea | 11:09 |
@wiking | wiking distcc | 11:09 |
lisitsyn | no | 11:09 |
@wiking | and i was about to kill u | 11:09 |
lisitsyn | you get me wrong | 11:09 |
@wiking | lisitsyn, you can actually | 11:09 |
@wiking | use for this buildbot | 11:09 |
@wiking | http://docs.buildbot.net/current/manual/cmdline.html#try | 11:10 |
@wiking | :) | 11:10 |
@wiking | Heiko is always crying about this feature | 11:10 |
@wiking | i told him to read the manual | 11:10 |
@wiking | and that's all | 11:10 |
@wiking | but he is too fucking lazy | 11:10 |
lisitsyn | wiking: I am just curious if it is possible to employ distcc somehow | 11:10 |
@wiking | it is | 11:10 |
@wiking | but there's buildbot | 11:10 |
lisitsyn | well in theory | 11:10 |
@wiking | it would do it for you | 11:10 |
@wiking | i mean everything is ready for that | 11:11 |
@wiking | you just need to actually type in the right command | 11:11 |
@wiking | :DDD | 11:11 |
@wiking | y do what distcc does it for ya | 11:11 |
lisitsyn | uhm not really | 11:11 |
@wiking | why not? | 11:11 |
lisitsyn | it's like you're ready to push | 11:11 |
lisitsyn | but you want to test again | 11:11 |
@wiking | nono | 11:11 |
@wiking | you can use a feature branch | 11:11 |
@wiking | and do all your experimentation there | 11:12 |
lisitsyn | no I mean it is not in the loop right? | 11:12 |
@wiking | which loop? | 11:12 |
lisitsyn | you just upload your code and test it somewhere | 11:12 |
lisitsyn | but with distcc you compile the same thing faster but test locally | 11:12 |
lisitsyn | right? | 11:12 |
@wiking | mmm | 11:12 |
@wiking | hahah | 11:13 |
@wiking | i wanna see the day | 11:13 |
@wiking | when we have that many macs | 11:13 |
@wiking | that you can distcc it | 11:13 |
@wiking | and then get it back to you | 11:13 |
@wiking | :))) | 11:13 |
@wiking | see the irony | 11:13 |
lisitsyn | wiking: ?? you don't need mac for that | 11:13 |
lisitsyn | you can cross compile | 11:13 |
@wiking | we dont even have one fucking mac builder | 11:13 |
@wiking | hahaha | 11:13 |
@wiking | have you done xcompile | 11:13 |
@wiking | :) | 11:13 |
@wiking | like ever :) | 11:13 |
lisitsyn | well with clang it works | 11:13 |
lisitsyn | yes we have it here | 11:13 |
@wiking | coz it's like maaaan sloow | 11:14 |
@wiking | compared to host compile | 11:14 |
lisitsyn | ok that's different issue | 11:14 |
@wiking | so i mean that's why | 11:14 |
@wiking | you would be just better of | 11:14 |
@wiking | fuckign sending your patch to buildbot | 11:14 |
@wiking | and it'll moan | 11:14 |
@wiking | if something is wrong | 11:14 |
lisitsyn | oh well then it's much easier to develop on some ssd server | 11:14 |
@wiking | but i mean | 11:14 |
@wiking | hedonismbot is pretty fast | 11:14 |
lisitsyn | with yuuuge memory | 11:14 |
lisitsyn | yes | 11:14 |
@wiking | :) | 11:15 |
@wiking | especially if you ccache | 11:15 |
@wiking | i mean bigger stuff i test on that directly | 11:15 |
@wiking | or quick patches even sometimes | 11:15 |
@wiking | becuase till i get shit compiled locally | 11:15 |
@wiking | .... | 11:15 |
@wiking | so yeah | 11:15 |
@wiking | use a server | 11:15 |
@wiking | that's my personal take on that | 11:15 |
@wiking | lisitsyn, btw is it still compiling? | 11:16 |
@wiking | because then i guess it's in better state than it used to be or? | 11:16 |
lisitsyn | wiking: just restarted it | 11:16 |
lisitsyn | it is going ok yet | 11:16 |
@wiking | k | 11:16 |
@wiking | it should | 11:16 |
@wiking | the most problem lately i had | 11:17 |
@wiking | was with viennacl | 11:17 |
@wiking | and macbook pro's fucking gpu | 11:17 |
lisitsyn | yeah I just wasn't aware it could be that config.h stuff | 11:17 |
@wiking | that does not support double precision float :D | 11:17 |
@wiking | i almost had 5.0.0 merged into brew | 11:18 |
@wiking | but then some stupid stuff came | 11:18 |
@wiking | and then yeah | 11:18 |
@wiking | everything escalated | 11:18 |
@wiking | we merged windows | 11:18 |
@wiking | then parallel fix | 11:18 |
@wiking | then lalalal | 11:18 |
@wiking | and now we already have like linalg | 11:18 |
@wiking | so i dont have a fucking clue where are we at atm | 11:18 |
@wiking | :) | 11:18 |
@wiking | but like a week ago i think we had most of the unit tests passing | 11:18 |
@wiking | ok i'm out | 11:18 |
@wiking | bbl | 11:19 |
lisitsyn | heh | 11:19 |
lisitsyn | ok | 11:19 |
lisitsyn | wiking: yes it worked | 11:19 |
lisitsyn | gracias amigo | 11:19 |
iRmantou_ | I see the DEV build failing in my pr and the error message is " CMake 3.1 or higher is required"... It's my code problem? | 11:31 |
@lambday|work | holy shit man what happened here? | 11:44 |
CaBa | iRmantou_: check only travis | 11:49 |
CaBa | morning people | 11:49 |
iRmantou_ | It's evening for me... | 11:58 |
-!- neha___ [7ab11852@gateway/web/freenode/ip.122.177.24.82] has joined #shogun | 12:11 | |
-!- rads [8984fa08@gateway/web/freenode/ip.137.132.250.8] has joined #shogun | 12:32 | |
lisitsyn | lambday|work: what? | 12:36 |
-!- iRmantou_ [55cb2405@gateway/web/freenode/ip.85.203.36.5] has quit [Ping timeout: 260 seconds] | 12:37 | |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has joined #shogun | 13:08 | |
@lambday|work | lisitsyn: really long logs :D | 13:09 |
@lambday|work | lisitsyn: I have a question | 13:09 |
@lambday|work | lisitsyn: suppose, I want to observe method internal variables.. not class variables.. what would be a neat way to do that | 13:09 |
@lambday|work | let me give you an actual use-case | 13:10 |
@lambday|work | lisitsyn: https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/statistical_testing/QuadraticTimeMMD.h#L268 | 13:11 |
lisitsyn | lambday|work: I think one may implement observable<T> | 13:12 |
@lambday|work | so, permutation indices are something that the method generates in every iteration.. | 13:12 |
@lambday|work | lisitsyn: can you elaborate a bit more | 13:12 |
-!- abhinav_ [~abhinav@103.211.53.10] has joined #shogun | 13:12 | |
@lambday|work | what's the idea | 13:12 |
@lambday|work | lisitsyn: I need to be able to turn it on/off .. with some unifying API.. not these CQuadraticTimeMMD::save_permutation_inds() and then CQuadraticTimeMMD::get_permutation_inds()... | 13:15 |
lisitsyn | lambday|work: I am a bit afraid it is like debugger | 13:15 |
@lambday|work | lisitsyn: yeah but I'd want to access these values from modular interfaces... so plain old debug messages won't really do | 13:17 |
@lambday|work | observable<T> vars should be retrieved | 13:17 |
-!- rads [8984fa08@gateway/web/freenode/ip.137.132.250.8] has quit [Quit: Page closed] | 13:19 | |
-!- rads [8984fa08@gateway/web/freenode/ip.137.132.250.8] has joined #shogun | 13:20 | |
-!- travis-ci [~travis-ci@54.81.191.55] has joined #shogun | 13:22 | |
travis-ci | it'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/213766343 | 13:22 |
-!- travis-ci [~travis-ci@54.81.191.55] has left #shogun [] | 13:22 | |
-!- Killin [~Killin-A1@2602:304:68a6:ddc0:1c90:463:a576:3640] has quit [Ping timeout: 246 seconds] | 13:23 | |
@wiking | lisitsyn, ok so it was actually only that you had trash in your source :) | 13:30 |
lisitsyn | wiking: yes exactly | 13:30 |
lisitsyn | well you fixed it | 13:30 |
@wiking | i mean i broke it in the first place :) | 13:30 |
@wiking | so easy to fix shit like that :D | 13:31 |
@wiking | shogun-buildbot: force build 'release xenial deb' | 13:32 |
shogun-buildbot | build #0 forced | 13:32 |
shogun-buildbot | I'll give a shout when the build finishes | 13:32 |
@wiking | now let's see if this falls apart :))) | 13:32 |
shogun-buildbot | Hey! build release xenial deb #0 is complete: Failure [failed] | 13:32 |
shogun-buildbot | Build details are at http://buildbot.shogun-toolbox.org/builders/release%20xenial%20deb/builds/0 | 13:32 |
@wiking | hahah | 13:33 |
@wiking | :) | 13:33 |
@wiking | now this is great | 13:33 |
@wiking | heiko | 13:33 |
@wiking | Cloning into 'doc/wiki'... | 13:33 |
@wiking | fatal: remote error: | 13:33 |
@wiking | Repository not found. | 13:33 |
@wiking | fatal: clone of 'git://github.com/shogun-toolbox/shogun-wiki.git' into submodule path 'doc/wiki' failed | 13:33 |
@wiking | YOU JUST FUCKING BROKE STABLE BRANCH | 13:33 |
@wiking | how great is that | 13:33 |
@wiking | :> | 13:33 |
-!- Adi__ [67e69bda@gateway/web/freenode/ip.103.230.155.218] has joined #shogun | 13:34 | |
-!- Adi__ [67e69bda@gateway/web/freenode/ip.103.230.155.218] has quit [Client Quit] | 13:34 | |
@wiking | i'm wondering if we gonna get this software engineering as a job | 13:36 |
-!- neha___ [7ab11852@gateway/web/freenode/ip.122.177.24.82] has quit [Ping timeout: 260 seconds] | 13:36 | |
@wiking | right :) | 13:36 |
@wiking | or we gonna be like mad scientist doing shit all the time :))) | 13:36 |
@wiking | and randomly delete repositories | 13:36 |
@wiking | so nothing could be recreated :) | 13:36 |
@wiking | "NOTE: Currently, upgrading to new versions requires a fresh install. | 13:38 |
@wiking | " | 13:38 |
@wiking | these type of stuff :))) | 13:38 |
@wiking | amazing | 13:38 |
@wiking | lisitsyn, so tell me more ?:) | 13:38 |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 13:41 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 13:41 | |
@wiking | shogun-buildbot: force build 'release xenial deb' | 13:42 |
shogun-buildbot | build #1 forced | 13:42 |
shogun-buildbot | I'll give a shout when the build finishes | 13:42 |
shogun-buildbot | Hey! build release xenial deb #1 is complete: Failure [failed changelog] | 13:42 |
shogun-buildbot | Build details are at http://buildbot.shogun-toolbox.org/builders/release%20xenial%20deb/builds/1 | 13:42 |
@wiking | let's hope i've managed to trick it | 13:42 |
@wiking | woohooo not really | 13:42 |
lisitsyn | wiking: tell what? | 13:43 |
@wiking | lisitsyn, about engineering :) | 13:44 |
@wiking | man i'm tired :D | 13:44 |
lisitsyn | uhmm I wasn't following | 13:44 |
@wiking | noooooooooi cannot do what i though i can | 13:44 |
@wiking | :)) | 13:44 |
@wiking | lisitsyn, nothing...just being tired of not getting anything right | 13:44 |
@wiking | with this fucking project called shogun | 13:44 |
@wiking | :( | 13:44 |
lisitsyn | you have to suffer to thrive | 13:45 |
lisitsyn | :D | 13:45 |
@wiking | lol | 13:45 |
-!- shogun-buildbot [~shogun-bu@7nn.de] has quit [Quit: buildmaster reconfigured: bot disconnecting] | 13:47 | |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 13:47 | |
CaBa | wiking: heya | 13:55 |
-!- abhinav_ [~abhinav@103.211.53.10] has quit [Read error: Connection reset by peer] | 13:55 | |
@wiking | y0 | 13:55 |
-!- abhinav_ [~abhinav@103.211.53.10] has joined #shogun | 13:56 | |
@wiking | CaBa, sup ? :))) | 13:59 |
CaBa | wiking: too much ;) | 13:59 |
@wiking | :DDDDDDDDDDDDD | 13:59 |
@wiking | coo | 14:00 |
@wiking | cool | 14:00 |
@wiking | me too | 14:00 |
@wiking | :) | 14:00 |
@wiking | let's do | 14:00 |
@wiking | git rm -rf * | 14:00 |
-!- rads [8984fa08@gateway/web/freenode/ip.137.132.250.8] has quit [Ping timeout: 260 seconds] | 14:00 | |
@wiking | git commit . | 14:00 |
@wiking | git push | 14:00 |
@wiking | ? | 14:00 |
@wiking | what about that? :D | 14:00 |
CaBa | wiking: i like! | 14:00 |
@wiking | \o/ | 14:00 |
@wiking | let's get shit real | 14:00 |
@wiking | btw honestly | 14:00 |
@wiking | DynArray | 14:00 |
@wiking | DynamicObject lalala | 14:00 |
lisitsyn | TOTAL BULLSHIT | 14:00 |
@wiking | that shoudl be in the trash | 14:00 |
@wiking | :DDDDDDDDDD | 14:00 |
@wiking | lisitsyn, right? | 14:00 |
lisitsyn | yes | 14:00 |
CaBa | wiking: guess what a lough i had when i discovered the "WRAPPED" brother that is only compiled when metaexamples are compiled or some shit like that ^^ | 14:01 |
@wiking | yeah dont even go there | 14:04 |
@wiking | lisitsyn, so we through out? :) | 14:04 |
@wiking | and use STL? | 14:04 |
@wiking | or have some better idea? | 14:04 |
@wiking | CaBa, oh man those wrapped thins | 14:04 |
@wiking | *things | 14:04 |
@wiking | that's some serious hardcode stuff | 14:05 |
@wiking | you know when you fix a thing | 14:05 |
@wiking | with changing the code | 14:05 |
CaBa | shogun is full of hardcore | 14:05 |
@wiking | until it doesn't compile :D | 14:05 |
@wiking | and once it's ready | 14:05 |
@wiking | then you are like | 14:05 |
@wiking | "dont touch it, it works" | 14:05 |
@wiking | anyhow | 14:05 |
@wiking | i'd be more than happy | 14:05 |
@wiking | throwing out that whole array object story | 14:06 |
@wiking | because it's a batshit crazy stuff | 14:06 |
CaBa | anyways i'll continue with the PR for now, it actually breaks stuff right now | 14:06 |
@wiking | \o/ | 14:07 |
@wiking | so why dont we fucking throw it out | 14:07 |
@wiking | :D | 14:07 |
CaBa | so lets get rid of that hard wired shit for now and then you guys can go for the big STL refactor after :P | 14:07 |
@wiking | lambday|work, ^ | 14:07 |
CaBa | need drugs. | 14:10 |
@wiking | ahhahaha | 14:12 |
@wiking | which ones? :) | 14:12 |
@wiking | btw note the channel is logged and the internet archive is limitless :D | 14:12 |
CaBa | wiking: well then better don't start selling now :P | 14:13 |
@wiking | :DDDDDDDDDDDDDDD | 14:14 |
@wiking | https://motherboard.vice.com/en_us/article/why-american-farmers-are-hacking-their-tractors-with-ukrainian-firmware | 14:48 |
@wiking | :> | 14:48 |
@wiking | keygen for a tractor :DDD | 14:49 |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 14:57 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 15:01 | |
-!- goksinen [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 15:01 | |
-!- rads [3a92880d@gateway/web/freenode/ip.58.146.136.13] has joined #shogun | 15:09 | |
CaBa | wiking: hm, different community. no midi music and standard window decoration ^^ | 15:09 |
-!- goksinen is now known as magicwund | 15:09 | |
-!- abhinav_ [~abhinav@103.211.53.10] has quit [Ping timeout: 240 seconds] | 15:11 | |
@sukey | Issue #3741 "Systematically test all machine train and apply multithreaded" opened by karlnapf - https://github.com/shogun-toolbox/shogun/issues/3741 | 15:14 |
@sukey | Issue #3741 "Systematically test all machine train and apply multithreaded" karlnapf added label: "entrance" - https://github.com/shogun-toolbox/shogun/issues/3741 | 15:14 |
CaBa | [...] Kluthe, for example, uses pig manure to power his tractor, which requires engine modifications that would likely violate John Deere's terms of service on newer machines. [...] | 15:16 |
CaBa | O.o | 15:16 |
-!- geektoni [c1cdd24c@gateway/web/freenode/ip.193.205.210.76] has joined #shogun | 15:17 | |
-!- lambday|work [945742af@gateway/web/freenode/ip.148.87.66.175] has quit [Quit: Page closed] | 15:26 | |
-!- HeikoS [~heiko@128.40.177.10] has joined #shogun | 15:31 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 15:31 | |
-!- Killin [~Killin-A1@129.7.0.237] has joined #shogun | 15:38 | |
-!- Olivier [81615ad9@gateway/web/freenode/ip.129.97.90.217] has joined #shogun | 15:44 | |
-!- Olivier is now known as Guest68523 | 15:45 | |
-!- Guest68523 [81615ad9@gateway/web/freenode/ip.129.97.90.217] has quit [Client Quit] | 15:45 | |
-!- olinguyen [81615ad9@gateway/web/freenode/ip.129.97.90.217] has joined #shogun | 15:46 | |
-!- nagdev [dce39546@gateway/web/freenode/ip.220.227.149.70] has quit [Ping timeout: 260 seconds] | 15:52 | |
-!- Killin [~Killin-A1@129.7.0.237] has quit [Ping timeout: 240 seconds] | 15:56 | |
-!- geektoni [c1cdd24c@gateway/web/freenode/ip.193.205.210.76] has quit [Ping timeout: 260 seconds] | 15:56 | |
olinguyen | Hey! I'm planning of applying for GSoC with a data project (application). If I wanted to know if it was possible to do more advanced deep learning with Shogun. I know there's a module for neural networks, but how hard would it be to do stuff like LSTMs, RNN, etc? | 15:59 |
@wiking | mmmm | 16:00 |
@wiking | the only way to judge a puding is to actually try it | 16:00 |
@wiking | :) | 16:00 |
@wiking | olinguyen, welcome! | 16:01 |
@wiking | and cool thing! | 16:01 |
@wiking | i suggest you to check into the NN part of shogun | 16:01 |
@wiking | it has a lot of room for improvement | 16:01 |
@wiking | and as well if you are seriously considering an application | 16:01 |
@wiking | please start reading about how to get started | 16:01 |
@wiking | and start sending in fixes for entrance tasks | 16:01 |
@wiking | as that is a hard requirement for a successful gsoc application | 16:02 |
olinguyen | ok, thanks for the pointers :)! | 16:03 |
@wiking | lisitsyn, bazel installed on hedonism | 16:03 |
@wiking | olinguyen, more than welcome | 16:03 |
@wiking | :) | 16:03 |
@wiking | olinguyen, use mailing list as well.... these type of information are usually circulating there as well | 16:03 |
olinguyen | will do! | 16:03 |
rads | hi! I'm exploring shogun as well! but the mailing list link didn't work for me.. where is it?? | 16:05 |
@wiking | rads yeah there's a tiny problem with it | 16:05 |
@wiking | namely archive is only existing | 16:06 |
@wiking | once you subscribe | 16:06 |
@wiking | yeah we are working on it | 16:06 |
@wiking | but sorry at the moment to get actual mails | 16:06 |
@wiking | you need to subscribe | 16:06 |
@wiking | :( | 16:06 |
rads | ohh okay lemme try that | 16:06 |
rads | I attended the FOSSASIA shogun talk btw!! | 16:07 |
@wiking | haha | 16:12 |
@wiking | cool | 16:12 |
@wiking | you can tell then about it to HeikoS | 16:12 |
@wiking | :DDDD | 16:12 |
@HeikoS | :D | 16:13 |
@wiking | rads, so did you like it? :D | 16:13 |
@wiking | if you didn't no GSOC for you | 16:13 |
@wiking | :P | 16:13 |
@wiking | joking of course | 16:13 |
@wiking | please share | 16:13 |
rads | yes, in fact i talked to you after it | 16:13 |
rads | :P | 16:13 |
@wiking | ah cool | 16:13 |
@wiking | then i know who you are | 16:13 |
@wiking | :) | 16:13 |
rads | about the kaggle one :P | 16:14 |
@wiking | great | 16:14 |
@wiking | and welcome to shogun | 16:14 |
-!- lambday [6a33177c@gateway/web/freenode/ip.106.51.23.124] has joined #shogun | 16:15 | |
-!- mode/#shogun [+o lambday] by ChanServ | 16:15 | |
rads | thanks! :D | 16:16 |
@lambday | guys.. where can I read the logs :| | 16:16 |
@lambday | none of the links work | 16:16 |
@wiking | :))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))0 | 16:16 |
@wiking | this is soooooooooo good | 16:16 |
@wiking | that you came with this question now | 16:16 |
@wiking | :> | 16:16 |
@lambday | wiking: who, me? | 16:17 |
@wiking | yes | 16:17 |
@wiking | you | 16:17 |
@wiking | ;) | 16:17 |
@HeikoS | logs of channel? | 16:17 |
@lambday | yessir | 16:17 |
@lambday | just try clicking the above two links that we read in the channel header | 16:17 |
@HeikoS | ah yes they stopped working | 16:17 |
@HeikoS | http://shogun.ml/irclogs | 16:17 |
@lambday | HeikoS: try this : http://shogun.ml/irclog/2017-03-22/ | 16:18 |
@lambday | I am getting Internal Server Error The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application. | 16:18 |
@HeikoS | same here | 16:18 |
@HeikoS | no idea what happened there | 16:18 |
@lambday | :| | 16:18 |
@lambday | this is so bad | 16:18 |
@HeikoS | I was aware of it but not sure whats going on | 16:18 |
@wiking | how was that saying? | 16:18 |
@lambday | so many good discussion.. all gone :'( | 16:18 |
@HeikoS | hahaha | 16:19 |
@wiking | when maradona handed in that | 16:19 |
@wiking | goal | 16:19 |
@wiking | in 86? | 16:19 |
@wiking | god's magic hand? :d | 16:19 |
@lambday | hands of god? | 16:19 |
@wiking | btw | 16:19 |
@wiking | i can confirm | 16:19 |
@wiking | it's not gon | 16:19 |
@wiking | *gone | 16:19 |
@wiking | :) | 16:19 |
@wiking | the logs are there | 16:19 |
@lambday | wiking: how can I access it | 16:19 |
@wiking | honestly? :) | 16:19 |
@lambday | wiking: honestly ;) | 16:20 |
@wiking | i can see it via the shell | 16:20 |
-!- HeikoS [~heiko@128.40.177.10] has quit [Remote host closed the connection] | 16:20 | |
@lambday | what shell | 16:20 |
@wiking | dunno why the web broke | 16:20 |
@wiking | the shell to the server | 16:20 |
@wiking | where the web is running atm | 16:20 |
@lambday | I don't have access | 16:20 |
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun | 16:20 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:20 | |
@wiking | yeah that i know | 16:21 |
@wiking | :) | 16:21 |
@wiking | i can zip it for you | 16:21 |
@wiking | :) | 16:21 |
@lambday | HeikoS: hey did you follow our (a) immutable features (b) subsets via operators discussion? | 16:21 |
@HeikoS | neither of that | 16:21 |
@HeikoS | lambday: but +1 for immutable features | 16:21 |
@HeikoS | subsets are a bit more tricky | 16:21 |
@lambday | HeikoS: subsets --> linear operators, right? | 16:21 |
@lambday | dimension subsets too | 16:21 |
@HeikoS | sure but then we need to rewrite the algos in terms of that API | 16:22 |
@HeikoS | so a bit of work | 16:22 |
@HeikoS | but algos should work against API anyways, not matrix | 16:22 |
@lambday | HeikoS: but imagine how clean our features API would be | 16:22 |
@lambday | exactly | 16:22 |
@HeikoS | masking for subsets | 16:22 |
@lambday | that's internal part.. | 16:22 |
@HeikoS | but only for subset | 16:22 |
@lambday | impl details | 16:22 |
-!- neha [ca8e6c97@gateway/web/freenode/ip.202.142.108.151] has joined #shogun | 16:22 | |
@HeikoS | for masking dimensions, it is harder | 16:22 |
@HeikoS | as dot needs to be overwritten | 16:22 |
@HeikoS | but yeah | 16:23 |
@HeikoS | that is a cool thing to do | 16:23 |
-!- Killin [~Killin-A1@129.7.0.237] has joined #shogun | 16:23 | |
@lambday | how about we do it the mathematical way... vector features ... provides (a) add (b) scale... | 16:23 |
@lambday | normed... specifies norm | 16:23 |
@lambday | inner prod\ | 16:23 |
@lambday | like that | 16:23 |
@lambday | algorithms simply rely on these functionalities | 16:23 |
@HeikoS | lambday: yeah it is like algos are coded against features API | 16:24 |
@HeikoS | not linalg | 16:24 |
@lambday | HeikoS: that's how it should be, no? | 16:24 |
@HeikoS | yep | 16:24 |
@lambday | linalg is just one special case | 16:24 |
@HeikoS | and oop polymorphism decides which things work and which dont | 16:24 |
@HeikoS | but for that we need some cleaner API | 16:25 |
@lambday | HeikoS: that's the idea | 16:25 |
@HeikoS | and java like interfaces or something | 16:25 |
@HeikoS | so this stuff will all be internal API | 16:25 |
@HeikoS | not SWIG | 16:25 |
@lambday | HeikoS: without MI, we can do that using managers | 16:25 |
@HeikoS | why not do MI? | 16:26 |
@HeikoS | I never got that | 16:26 |
@HeikoS | such a good thing, classes with only purely virutal methods | 16:26 |
@lambday | HeikoS: we may run into weird errors | 16:26 |
@HeikoS | but yeah can also do managers, i am just less familiar with that | 16:26 |
@HeikoS | used java interfaces for ages for such things | 16:26 |
@lambday | HeikoS: so you're saying, use MI for internal structure.. not for SWIG wrappers | 16:27 |
@HeikoS | yeah | 16:29 |
@HeikoS | build a proper ML lib interface inside | 16:29 |
@HeikoS | and then expose gently | 16:29 |
@HeikoS | but there are problems with this as well | 16:29 |
@HeikoS | we dont want to write new interface classes for every algorithm | 16:29 |
@HeikoS | BUT | 16:29 |
@HeikoS | with the plugins stuff | 16:29 |
@HeikoS | we will only have BASE ML interface to swig anyways | 16:29 |
@HeikoS | and the plugins can do their own thing inside | 16:29 |
@HeikoS | so can libshogun | 16:30 |
@HeikoS | so we can do proper stuff there | 16:30 |
@HeikoS | tea time! | 16:30 |
@HeikoS | ill be back later | 16:31 |
@wiking | HeikoS, take care | 16:31 |
@HeikoS | shall we discuss more then? | 16:31 |
@wiking | HeikoS, there's shit storm in london atm | 16:31 |
@lambday | HeikoS: sure | 16:31 |
@HeikoS | wiking: shit storm? | 16:31 |
@wiking | HeikoS, hope you are far away from it | 16:31 |
@wiking | westminster bridge | 16:31 |
CaBa | more like... bullets... | 16:31 |
@wiking | if you dont know | 16:31 |
@wiking | then read bbc | 16:31 |
@lambday | wiking: wtf dude | 16:31 |
@HeikoS | oh wow | 16:31 |
@HeikoS | checking | 16:31 |
-!- ikazu [a027de6e@gateway/web/freenode/ip.160.39.222.110] has joined #shogun | 16:31 | |
CaBa | :( | 16:32 |
@lambday | wiking: umm.. if things are internal, can't we use multiple template params instead of MI? | 16:34 |
@lambday | just thinking out loud | 16:35 |
@wiking | eh? | 16:35 |
@wiking | havne't been following | 16:35 |
@wiking | :) | 16:35 |
@sukey | Pull Request #3721 "Alternative fix for Dynamic(Object)Array::clone()" synchronized by lkuchenb - https://github.com/shogun-toolbox/shogun/pull/3721 | 16:35 |
@lambday | wiking: let me scribble something | 16:36 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 16:37 | |
CaBa | HeikoS: sorry i vaguely remember i asked that already, but what was the issue with init list vs register_params() assign? | 16:39 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 16:39 | |
CaBa | HeikoS: and then... just a day ago or so you said you wanted indent / whitespace corrections in separate commits and now you're telling me in #3731 to squash? | 16:42 |
-!- neha [ca8e6c97@gateway/web/freenode/ip.202.142.108.151] has quit [Ping timeout: 260 seconds] | 16:45 | |
@HeikoS | CaBa: jojo | 16:47 |
CaBa | heya | 16:47 |
@HeikoS | CaBa: so whitespace -> you are right | 16:47 |
@HeikoS | maybe I did see something wrong, but definitely in sep commits | 16:47 |
@HeikoS | CaBa: in #3731 there are just a few more of those doxy comments that you copied in | 16:48 |
CaBa | yep i know, killing them | 16:48 |
CaBa | i'm more wondering about the init point | 16:48 |
@HeikoS | CaBa: yeah that thing | 16:49 |
@HeikoS | so registered parameters should get initialised in the same method as they are registered | 16:49 |
CaBa | because? | 16:49 |
@HeikoS | otherwise, it can be forgotten to be initialized at all (muliple constructors) and this has lead to errors earlier | 16:49 |
CaBa | k. any pref as to init before or after reg? | 16:50 |
@HeikoS | doesnt matter, just in the same method | 16:50 |
CaBa | k | 16:50 |
@HeikoS | and then this same method is called from all constructors | 16:50 |
@HeikoS | which means you cannot use the init fields, but have to assign by hand | 16:50 |
CaBa | well there is only one constructur :P | 16:50 |
CaBa | but who knows... more may come in the future ^^ | 16:50 |
@HeikoS | sure, it is more style of Shogun, we should keep it consistent | 16:51 |
@HeikoS | lambday: soso | 16:51 |
@lambday | HeikoS: yo | 16:51 |
@HeikoS | lambday: how to proceed with these features | 16:51 |
@lambday | HeikoS: trying to scribble something | 16:51 |
@HeikoS | making everything read only would be a start? | 16:52 |
@lambday | HeikoS: sure | 16:52 |
@lambday | HeikoS: but in order to do that, we gotta move subsets out | 16:52 |
@lambday | that itself is a massive task | 16:52 |
@HeikoS | why? | 16:52 |
@lambday | HeikoS: let's have the design on paper first | 16:52 |
@HeikoS | yeah ok | 16:52 |
@lambday | we gotta make sure that we thought this through :D | 16:53 |
CaBa | HeikoS: you have a standard name for such a func? register_params is only half correct now ;) | 16:53 |
@HeikoS | we used to call it "init" | 16:53 |
@HeikoS | or "register_params" | 16:54 |
-!- Killin [~Killin-A1@129.7.0.237] has quit [Ping timeout: 240 seconds] | 16:54 | |
CaBa | HeikoS: access specifier? | 16:54 |
@HeikoS | ? | 16:55 |
CaBa | protected or private? | 16:55 |
@HeikoS | private | 16:55 |
@HeikoS | it is never called from elsewhere | 16:55 |
@HeikoS | subclass calls baseclass constructor | 16:55 |
lisitsyn | stupid q | 16:57 |
lisitsyn | does sgvector copy on assignment? :D | 16:58 |
lisitsyn | rat shit bat shit | 16:59 |
@HeikoS | lisitsyn: didnt you make that? | 16:59 |
lisitsyn | make what | 17:00 |
@HeikoS | copy | 17:00 |
lisitsyn | no idea | 17:00 |
lisitsyn | I can't remember all the evil I made | 17:01 |
@HeikoS | we did a lot of const ref changes last year when vectors are passed | 17:01 |
@wiking | btw | 17:02 |
@wiking | how about | 17:02 |
@wiking | this is just a mental note here | 17:02 |
@wiking | i truely dont wanna shit over this whole effort of making things better | 17:02 |
@wiking | but | 17:02 |
@wiking | SG_REF & SG_UNREF | 17:02 |
@wiking | :) | 17:03 |
lisitsyn | what's with sgref | 17:03 |
@wiking | i mean | 17:03 |
@HeikoS | it should go away | 17:03 |
@wiking | that is bat shit crazy | 17:03 |
@wiking | in 2017 | 17:03 |
lisitsyn | ah | 17:03 |
lisitsyn | do you see anyone against that? | 17:03 |
lisitsyn | :P | 17:03 |
@wiking | no i mean | 17:03 |
@wiking | what about putting significant | 17:04 |
@wiking | effort to do that | 17:04 |
@HeikoS | yeah +1 | 17:04 |
@wiking | i reckon that is a huge usability impact | 17:04 |
@wiking | and now that we *require* c++11 | 17:04 |
@HeikoS | I think: hackathon we make a draft and figure out details, GSoC a student deploys these | 17:04 |
@HeikoS | end of year, we integrate that stuff back in | 17:04 |
@wiking | we could even use like shared_ptr | 17:04 |
@wiking | or wrap it | 17:04 |
@wiking | i dont mind | 17:04 |
@wiking | just get rid of that logic | 17:05 |
@wiking | like asap | 17:05 |
lisitsyn | ehm we have a wrapper | 17:05 |
lisitsyn | :) | 17:05 |
@wiking | or say the least | 17:05 |
@wiking | some? | 17:05 |
@wiking | some still leaks | 17:05 |
lisitsyn | yes it is even used | 17:05 |
lisitsyn | really? | 17:05 |
@wiking | based on valgrind | 17:05 |
@wiking | :) | 17:05 |
@HeikoS | no some pls | 17:05 |
lisitsyn | why? | 17:05 |
@HeikoS | thats the meta examples | 17:05 |
@HeikoS | they leak | 17:05 |
@wiking | :D | 17:05 |
lisitsyn | well make me fix it | 17:05 |
@wiking | ok so but what then | 17:05 |
@HeikoS | because I stole sergeys wrap | 17:05 |
@wiking | lisitsyn, we asked | 17:05 |
@wiking | several time | 17:05 |
@wiking | s | 17:05 |
@wiking | but man your memory | 17:06 |
@wiking | is like 0 | 17:06 |
@wiking | :) | 17:06 |
lisitsyn | what memory | 17:06 |
lisitsyn | :D | 17:06 |
@wiking | :D | 17:06 |
@wiking | good | 17:06 |
@wiking | robot | 17:06 |
@wiking | :) | 17:06 |
@wiking | now dance ! :) | 17:06 |
lisitsyn | don't hurt me I am a poor russian | 17:06 |
@wiking | shogun-buildbot: dance | 17:06 |
shogun-buildbot | <(^.^<) | 17:06 |
shogun-buildbot | <(^.^)> | 17:06 |
shogun-buildbot | (>^.^)> | 17:06 |
shogun-buildbot | (7^.^)7 | 17:06 |
shogun-buildbot | (>^.^<) | 17:06 |
@wiking | sukey flip | 17:06 |
@sukey | (╯°□°)╯︵ ┻━┻ | 17:06 |
@wiking | :> | 17:06 |
@wiking | lisitsyn, there poor russian robot :) | 17:06 |
@wiking | here are your friends play with them | 17:07 |
@wiking | ok but yeah | 17:07 |
@wiking | i mean i'm fine of at least having something | 17:07 |
@wiking | that we require | 17:07 |
@wiking | all the gsoc students | 17:07 |
@wiking | to use | 17:07 |
lisitsyn | well set up a mail that forces me to force some | 17:07 |
@wiking | and whatever they encounter | 17:07 |
@wiking | they change to that | 17:07 |
lisitsyn | it is nothing but a shared_ptr | 17:07 |
@wiking | yeah i know | 17:07 |
lisitsyn | replacing it with shared_ptr won't fix that | 17:07 |
lisitsyn | because it is a shared_ptr | 17:07 |
lisitsyn | you bastards :D | 17:08 |
@wiking | mmm | 17:08 |
@wiking | shared_ptr is atomic | 17:08 |
@wiking | no? | 17:08 |
@wiking | is some atomic? | 17:08 |
@wiking | :) | 17:08 |
lisitsyn | so who has 0 memory now | 17:08 |
@wiking | ah i should know c++ standard | 17:08 |
@wiking | by heart? :) | 17:08 |
@wiking | that's the point here | 17:09 |
lisitsyn | I have no idea if it is atomic | 17:09 |
lisitsyn | it's as atomic as SG_REF | 17:09 |
@wiking | ref is | 17:09 |
@wiking | :) | 17:09 |
@sukey | New Commit "Proper way to handle W in OCAS SVM" to shogun-toolbox/shogun by lisitsyn: https://github.com/shogun-toolbox/shogun/commit/b7b9dfeb9a2640c8eede9e13e72e870167b12e9e | 17:10 |
@sukey | Pull Request #3674 "Use weight vector *only* through getters/setters" synchronized by lisitsyn - https://github.com/shogun-toolbox/shogun/pull/3674 | 17:10 |
lisitsyn | bitch please! | 17:10 |
@wiking | oooooo | 17:10 |
@wiking | sergey just commited | 17:10 |
@wiking | :) | 17:10 |
lisitsyn | ohoho | 17:10 |
lisitsyn | it's because you broke my life | 17:10 |
@wiking | ahhaha | 17:10 |
@wiking | i did i see | 17:10 |
lisitsyn | so you some-haters | 17:12 |
@wiking | but yeah | 17:12 |
lisitsyn | tell me how do you replace it | 17:12 |
@wiking | what the fuck we d with it | 17:12 |
lisitsyn | :P | 17:12 |
@wiking | because | 17:12 |
@wiking | now that we are going to be having | 17:12 |
lisitsyn | show me where does it leak | 17:12 |
@wiking | just another way to do | 17:12 |
@wiking | shit | 17:12 |
@wiking | is a bit crazy | 17:12 |
@wiking | lisitsyn, ask valgrind | 17:12 |
@wiking | what am i, a service? | 17:12 |
@wiking | :) | 17:12 |
@HeikoS | the meta example leak | 17:12 |
lisitsyn | $ valgrind [19:07:43] | 17:12 |
lisitsyn | valgrind: no program specified | 17:12 |
lisitsyn | valgrind: Use --help for more information. | 17:12 |
@HeikoS | but not because of some | 17:12 |
@HeikoS | but because of wrap | 17:12 |
@wiking | :> | 17:13 |
@wiking | sexy | 17:13 |
@wiking | :D | 17:13 |
lisitsyn | wiking: it doesn't work | 17:13 |
@wiking | some, any, wrap, ref | 17:13 |
@wiking | lisitsyn, fucker! | 17:13 |
lisitsyn | wiking: what should I do | 17:13 |
@wiking | lisitsyn, i knew that we cannot trust services!!! | 17:13 |
@wiking | microservices are all lying | 17:13 |
@wiking | !!! | 17:13 |
lisitsyn | valgrind doesn't work | 17:13 |
@wiking | aaaaaaaaaaaaaaaaaaaaaaaaahh | 17:13 |
-!- OXPHOS [92bd305b@gateway/web/freenode/ip.146.189.48.91] has quit [Ping timeout: 260 seconds] | 17:13 | |
@wiking | they are all just waiting to kill us | 17:13 |
@wiking | microwars | 17:13 |
@wiking | aaaaaaaaaaaaaaaaaaaa | 17:13 |
lisitsyn | val grind | 17:14 |
lisitsyn | hello | 17:14 |
@HeikoS | guys | 17:14 |
@HeikoS | guys | 17:14 |
@wiking | mmm yeah man that girnds | 17:14 |
@HeikoS | code of conduct | 17:14 |
@HeikoS | :D | 17:14 |
@wiking | where | 17:14 |
@wiking | ? | 17:14 |
@wiking | who? | 17:14 |
lisitsyn | code of valgrind | 17:14 |
@wiking | grind it! | 17:14 |
@wiking | GRIND IT DOWN! | 17:14 |
@wiking | GRIND IT DOWN! | 17:14 |
lisitsyn | I think I'll send the first complaint | 17:14 |
@wiking | do you think if we add some microservices to shogun | 17:14 |
lisitsyn | to this wiking | 17:14 |
@wiking | things will start working? | 17:14 |
@wiking | we need webscale! | 17:15 |
@wiking | maybe adding mongodb | 17:15 |
@wiking | will hep | 17:15 |
@wiking | *help | 17:15 |
lisitsyn | it should work | 17:15 |
@wiking | ok | 17:15 |
@wiking | i thought so | 17:15 |
@wiking | :) | 17:15 |
@wiking | so we add couple of microservices | 17:15 |
@wiking | and mongodb | 17:15 |
lisitsyn | wiking: http://buildbot.shogun-toolbox.org/builders/cookbook%20-%20PR/builds/524/steps/configure/logs/stdio | 17:15 |
@wiking | anything else? | 17:15 |
lisitsyn | fix it plz | 17:15 |
lisitsyn | plz | 17:15 |
lisitsyn | :D | 17:15 |
lisitsyn | devops | 17:15 |
@wiking | do we need redis? | 17:15 |
lisitsyn | plez | 17:15 |
@wiking | maybe redis as well | 17:15 |
lisitsyn | pleazee | 17:15 |
@wiking | UNFIXABLE | 17:16 |
@wiking | :) | 17:16 |
lisitsyn | I don't have the password | 17:16 |
@wiking | so redis, mongodb | 17:16 |
lisitsyn | can't touch this | 17:16 |
@wiking | lisitsyn, dont be a bithc please | 17:16 |
@wiking | you have access to shogun-infra | 17:16 |
lisitsyn | it doesn't work | 17:16 |
@wiking | there's the whole code for buildbot | 17:16 |
@wiking | you can move that stupid service | 17:16 |
@wiking | to any host | 17:16 |
@wiking | :) | 17:16 |
@wiking | so | 17:16 |
lisitsyn | but it doesn't work | 17:16 |
@wiking | open it with your little hands | 17:16 |
@wiking | and make it work | 17:16 |
@wiking | :) | 17:16 |
@wiking | microservices, no? | 17:16 |
@wiking | i think that should work | 17:17 |
lisitsyn | $ brew install redis | 17:17 |
lisitsyn | still | 17:17 |
lisitsyn | wiking: fix it pleaze | 17:17 |
@wiking | mmm | 17:17 |
@wiking | zookeeper | 17:17 |
@wiking | you need that as well | 17:17 |
@wiking | ! | 17:17 |
lisitsyn | ruok | 17:18 |
@wiking | yeyeyee! | 17:18 |
@wiking | good idea! | 17:18 |
lisitsyn | imok | 17:18 |
lisitsyn | wiking: ruok | 17:18 |
@wiking | but we need consistency | 17:18 |
lisitsyn | wiking: ruok!!!!! | 17:18 |
@wiking | so how do we do that?! | 17:18 |
@wiking | and a fast in memory db | 17:18 |
@wiking | i think that's rockdb | 17:18 |
@wiking | no? | 17:18 |
@wiking | we need that | 17:18 |
lisitsyn | good choice | 17:18 |
@wiking | because for us | 17:18 |
@wiking | using ramfs | 17:18 |
@wiking | is just too oldschoool | 17:18 |
@wiking | like for facebook | 17:18 |
@wiking | :) | 17:18 |
@wiking | lisitsyn, i hope you appreciate the irony :) | 17:19 |
@wiking | lisitsyn, somebody wrote an article about this story with rockdb | 17:19 |
lisitsyn | HeikoS: what's wrap | 17:19 |
lisitsyn | HeikoS: this guy just blames me can you point me a bit closer to where does it leak | 17:19 |
@HeikoS | yeah | 17:20 |
lisitsyn | wiking: oracle should work | 17:20 |
@HeikoS | we instantiate objects in meta example | 17:20 |
@HeikoS | check the c++ output | 17:20 |
@HeikoS | or no that is some | 17:20 |
@HeikoS | but then a SGOBject is returned | 17:20 |
@HeikoS | something like that | 17:20 |
@HeikoS | check c++ listing that is generated | 17:20 |
@HeikoS | or memcheck the tests | 17:21 |
lisitsyn | meta tests? | 17:21 |
@HeikoS | ctest -R generated_cpp-* | 17:21 |
lisitsyn | kk | 17:21 |
-!- OXPHOS [92bd305b@gateway/web/freenode/ip.146.189.48.91] has joined #shogun | 17:21 | |
@wiking | lisitsyn, mmmm do we want enterprise? | 17:21 |
@wiking | i mean ok if we want enterprise then we have to do oracle | 17:21 |
@wiking | lisitsyn, so ok | 17:24 |
@wiking | we do super microservices | 17:24 |
@wiking | every object is a service in fact! | 17:24 |
@wiking | ah fuck | 17:24 |
@wiking | !!! | 17:24 |
@wiking | we need RPC! | 17:24 |
lisitsyn | grpc | 17:24 |
@wiking | because you have to have that | 17:24 |
@wiking | !!! | 17:24 |
@wiking | yes yes | 17:24 |
@wiking | add that | 17:24 |
@wiking | good choice! | 17:24 |
-!- neha [ca8e6c97@gateway/web/freenode/ip.202.142.108.151] has joined #shogun | 17:24 | |
@wiking | mmm but one more thing | 17:26 |
@wiking | i believe | 17:26 |
@wiking | we need a heroku app | 17:26 |
@wiking | no? | 17:26 |
@HeikoS | we have one :D | 17:26 |
@wiking | and of course an ios and android app | 17:26 |
@wiking | DISCLAIMER: all this is a fiction now :) | 17:26 |
@wiking | before anybody starts to take this seriously | 17:26 |
@wiking | :) | 17:26 |
@wiking | lisitsyn, ok so how about we do a distributed db on every ios app | 17:27 |
@wiking | and connect them with grpc | 17:27 |
lisitsyn | wiking: thrift-grpc-bridge | 17:27 |
@wiking | ! | 17:27 |
@wiking | definitely! | 17:27 |
lisitsyn | with jackson-gson | 17:27 |
@wiking | good call! | 17:27 |
lisitsyn | but | 17:27 |
lisitsyn | most importantly | 17:27 |
lisitsyn | we run single thread | 17:28 |
lisitsyn | in python | 17:28 |
lisitsyn | because. | 17:28 |
lisitsyn | :D | 17:28 |
@wiking | GIL | 17:28 |
@wiking | of course | 17:28 |
@wiking | what else | 17:28 |
@wiking | there's no point of being without the GIL | 17:28 |
@wiking | lisitsyn, aaaaaaaaaaaah!!!! | 17:29 |
@wiking | asyncIO | 17:29 |
@wiking | only | 17:29 |
@wiking | ! | 17:29 |
@wiking | and of course with coroutines!!! | 17:29 |
@wiking | and if anybody asks we serialize *everything* into json | 17:30 |
@wiking | strictly | 17:30 |
@wiking | as binary formats are confusing | 17:30 |
lisitsyn | wiking: ok sorry to interrupt but I have another build issue | 17:30 |
lisitsyn | :D | 17:30 |
@wiking | :D | 17:30 |
@wiking | tell me | 17:31 |
lisitsyn | where should I run examples!!!!!!!! | 17:31 |
@wiking | well | 17:31 |
lisitsyn | data | 17:31 |
@wiking | if i tell it the way i want | 17:31 |
@wiking | i thnk i really gonna be getting some batshit crazy stuff | 17:31 |
@wiking | with the code of conduct | 17:31 |
lisitsyn | oh | 17:31 |
@wiking | so dont run it on your mama | 17:31 |
@wiking | :) | 17:31 |
lisitsyn | finally | 17:31 |
lisitsyn | found | 17:31 |
lisitsyn | wiking: ok continue please | 17:31 |
lisitsyn | :D | 17:31 |
@wiking | but ctest | 17:32 |
@wiking | that's all | 17:32 |
lisitsyn | nono | 17:32 |
@wiking | it'll run it for you | 17:32 |
@wiking | or you wanna valgrind it? | 17:32 |
lisitsyn | yes | 17:32 |
@wiking | ok | 17:32 |
lisitsyn | wiking: even though valgrind doesn't work! | 17:32 |
@wiking | then run it in the examples/undocumented/libshogun | 17:32 |
@wiking | lalal | 17:32 |
@wiking | because you know | 17:32 |
@wiking | WE ARE SO YOLO | 17:32 |
@wiking | that for us | 17:32 |
@wiking | ENV VARIABLE | 17:32 |
@wiking | is to derogating | 17:32 |
@wiking | we fucking hardcode paths | 17:32 |
@wiking | into the code | 17:33 |
@wiking | :) | 17:33 |
OXPHOS | wiking: can i jump in for a question? So for the xla backend, we're actually expecting a FindTensorflow.cmake, and everything should be pre-installed by users, right? | 17:33 |
@wiking | portability? only crazy people need that | 17:33 |
@wiking | :) | 17:33 |
@wiking | woooooooooooooooohooooooooooooooooooooo | 17:33 |
lisitsyn | ok some is broken | 17:33 |
@wiking | OXPHOS, just dopped th emic | 17:33 |
@wiking | *mic | 17:33 |
@wiking | OXPHOS, mmm the xla | 17:34 |
@wiking | you need actully the compilers | 17:34 |
OXPHOS | lol sorry for disruption | 17:34 |
@wiking | you dont need the tf | 17:34 |
@wiking | :) | 17:34 |
@wiking | no problem | 17:34 |
@wiking | no serious development is happening here | 17:34 |
lisitsyn | oh one more small dependency :D | 17:34 |
@wiking | right lisitsyn ?:) | 17:34 |
@HeikoS | OXPHOS: xla backend ... | 17:34 |
@HeikoS | needs some serious discussion | 17:34 |
@wiking | HeikoS, dont interrupt | 17:34 |
lisitsyn | la petite dependency! | 17:34 |
@wiking | dont need discussion | 17:34 |
@HeikoS | lol :) | 17:34 |
@wiking | shiiis | 17:34 |
@wiking | Pan is working | 17:34 |
@wiking | let her be! | 17:35 |
@HeikoS | OXPHOS: these guys got themselves into some serious excitement | 17:35 |
@wiking | OXPHOS, ok so | 17:35 |
OXPHOS | HeikoS: wiking really likes it | 17:35 |
@wiking | for xla you want find XLA | 17:35 |
@wiking | OXPHOS, you want me to write it for you? | 17:35 |
@HeikoS | me too | 17:35 |
OXPHOS | cool | 17:35 |
@HeikoS | but it wont work with the api we have | 17:35 |
OXPHOS | wiking: I did | 17:35 |
@wiking | because what u need | 17:35 |
@HeikoS | in linalg | 17:35 |
@wiking | is the JIT | 17:35 |
@wiking | HeikoS, could | 17:35 |
@wiking | it'll be a bit rudimentary | 17:35 |
@HeikoS | dont think so | 17:36 |
@wiking | but it could work | 17:36 |
@wiking | why not | 17:36 |
@wiking | i mena it wont work the most efficient way | 17:36 |
@wiking | but it is gonna work | 17:36 |
@HeikoS | because linalg blocks and gives you result | 17:36 |
@wiking | you just get the client | 17:36 |
@wiking | and do the blocks one by one | 17:36 |
@HeikoS | but in xla you draft first and then execute and wait for result | 17:36 |
@wiking | :) | 17:36 |
@HeikoS | is that a good idea? | 17:36 |
@wiking | yeah but who forces you | 17:36 |
@wiking | well | 17:36 |
@wiking | it's a start | 17:36 |
@wiking | no? | 17:36 |
@wiking | and then | 17:36 |
@wiking | later you can do nops | 17:36 |
@HeikoS | I think a different api is needed for that | 17:36 |
@wiking | and then execute it on the end | 17:36 |
@wiking | HeikoS, always getting lost in the details | 17:37 |
@wiking | look at lisitsyn | 17:37 |
@wiking | :) | 17:37 |
@wiking | OXPHOS, ok but so you realise | 17:37 |
lisitsyn | what | 17:37 |
@wiking | that xla is actually totally | 17:37 |
@wiking | detachable | 17:37 |
@wiking | from tf | 17:37 |
@HeikoS | linalg is more low level | 17:37 |
@wiking | right? | 17:37 |
@wiking | HeikoS, i mean from a codegen jit shit | 17:37 |
@wiking | you cannot | 17:37 |
@wiking | get more lowlevel | 17:37 |
@wiking | :) | 17:37 |
@HeikoS | sure | 17:37 |
@HeikoS | but the API | 17:37 |
OXPHOS | wiking: yes..but I didn't find a way, or haven't tried hard enough to find out how xla compile on itself | 17:38 |
@wiking | ah | 17:38 |
@wiking | ok | 17:38 |
@wiking | i se | 17:38 |
@wiking | HeikoS, yeah but that's ok | 17:38 |
@wiking | you can make xla work in pieces | 17:38 |
@HeikoS | would rather do GPU stuff and show its useful somewhere first | 17:38 |
@HeikoS | but hey we can work on it | 17:38 |
@HeikoS | its a cool topic | 17:38 |
@wiking | and then | 17:38 |
@HeikoS | yeah true | 17:38 |
@wiking | anyways | 17:38 |
@wiking | you are not forced | 17:38 |
@wiking | to execute | 17:38 |
@wiking | you can say | 17:38 |
@wiking | QR... dot ... allalala | 17:38 |
@HeikoS | sure, need a new API for tha tthough | 17:38 |
@wiking | and all these operations are stacked | 17:38 |
@HeikoS | linalg gives you SGMatrix in that case | 17:39 |
@wiking | and you return actually null matrices | 17:39 |
@wiking | :) | 17:39 |
@wiking | yeah and? | 17:39 |
@wiking | empty matrix | 17:39 |
@wiking | who cares | 17:39 |
@wiking | :) | 17:39 |
@HeikoS | and then you do something with the matrix | 17:39 |
CaBa | HeikoS: #3721, merge/ | 17:39 |
CaBa | ? | 17:39 |
@HeikoS | and do another linalg call | 17:39 |
@wiking | HeikoS, yeah but if its really only a series of linalgs | 17:39 |
@wiking | then you could do NOPs | 17:39 |
@wiking | i mean stack the oeprations | 17:39 |
@HeikoS | how does xla know this matrix is the one that was the qr result earlier? | 17:39 |
@wiking | and give back random matrix | 17:39 |
@HeikoS | as linalg takes a sgmatrix in | 17:39 |
@HeikoS | I get the output empty matrix thing | 17:39 |
@HeikoS | but how can you stack stuff up? | 17:40 |
@HeikoS | if it returns a sgmatrix? | 17:40 |
@wiking | if you dont touc | 17:40 |
@wiking | the output of the prev operation | 17:40 |
@wiking | then just call the next one | 17:40 |
@wiking | each time behind the linalg | 17:40 |
@wiking | it's the same client | 17:40 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 17:40 | |
@HeikoS | what if it returns multiple thing and you just want some of that? | 17:40 |
@wiking | and what if the sky is red and not blue? | 17:40 |
@wiking | dunno | 17:40 |
@HeikoS | i.e. API issues | 17:40 |
@wiking | but if we dont start doing this | 17:40 |
@HeikoS | linalg API is not suited | 17:41 |
@wiking | we never gonna do this | 17:41 |
@HeikoS | true that | 17:41 |
@wiking | and discussing this | 17:41 |
@wiking | is not going to help | 17:41 |
@HeikoS | but that needs something for graph computation | 17:41 |
@wiking | until it doesnt fucking break apart | 17:41 |
@HeikoS | like define computation before executring it | 17:41 |
@wiking | yeah cool | 17:41 |
@HeikoS | but hey | 17:41 |
@wiking | lets try to do it | 17:41 |
@wiking | and see where things start seriously | 17:41 |
@HeikoS | I already agreed :D | 17:41 |
@wiking | breaking apart | 17:41 |
@HeikoS | sure | 17:41 |
@HeikoS | next next hackathon :) | 17:41 |
@wiking | honestly | 17:41 |
@sukey | Pull Request #3730 "Register parameters for CMKL and override clone()" synchronized by lkuchenb - https://github.com/shogun-toolbox/shogun/pull/3730 | 17:41 |
@wiking | if we dont do this | 17:41 |
@wiking | we are fucked | 17:41 |
@HeikoS | yes thats the way to do things | 17:42 |
@wiking | since xla et al | 17:42 |
@wiking | is gonnna be maintained | 17:42 |
@wiking | and actually by a lot of people | 17:42 |
@HeikoS | yeah all ++ | 17:42 |
@wiking | our codebase? :) | 17:42 |
@wiking | SG_REF | 17:42 |
@wiking | :) | 17:42 |
@HeikoS | haha | 17:42 |
@wiking | i mean i guess you get the irony | 17:42 |
@wiking | :) | 17:42 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 17:43 | |
@HeikoS | /home/heiko/git/shogun/src/shogun/classifier/svm/SVMLight.cpp: In member function ‘float64_t* shogun::CSVMLight::optimize_qp(shogun::QP*, float64_t*, int32_t, float64_t*, int32_t&)’: | 17:43 |
@HeikoS | /home/heiko/git/shogun/src/shogun/classifier/svm/SVMLight.cpp:2444:9: error: ‘STILL_RUNNING’ was not declared in this scope | 17:43 |
@HeikoS | result=STILL_RUNNING; | 17:43 |
@HeikoS | config.h | 17:43 |
@HeikoS | changed branch | 17:43 |
@HeikoS | reran cmake | 17:43 |
@wiking | HeikoS, can you please | 17:43 |
@wiking | chack | 17:43 |
@wiking | that you | 17:43 |
@HeikoS | yet will fix it | 17:43 |
@wiking | dont have config.h, class_list.cpp, versionstring.h | 17:43 |
@wiking | in your SOURCE dirs | 17:44 |
@sukey | Pull Request #3731 "LibSVM parameters" synchronized by lkuchenb - https://github.com/shogun-toolbox/shogun/pull/3731 | 17:44 |
CaBa | so, guys, #3721 is ready for merge IMHO, #3730 and #3731 are in travis | 17:45 |
@wiking | see | 17:45 |
@wiking | another productive member here | 17:46 |
@wiking | :) | 17:46 |
OXPHOS | any conclusion :)? I'll probably get the xla and FinaXLA.cmake working for me first | 17:46 |
@wiking | OXPHOS, if u need help | 17:46 |
@wiking | lemme know | 17:47 |
@wiking | i'm gonna crash | 17:47 |
@wiking | but yeah | 17:47 |
@wiking | yo udont need tf | 17:47 |
@wiking | only XLA | 17:47 |
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Read error: Connection reset by peer] | 17:47 | |
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has joined #shogun | 17:47 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 17:47 | |
@wiking | OXPHOS, just try to install it with bazel | 17:47 |
@wiking | and then you'll get all the headers and libs and binaries in place | 17:48 |
OXPHOS | wiking: thanks. not sure what to ask yet. but I'll try | 17:48 |
OXPHOS | wiking: i did. and just find out you need extra steps to make JIT work | 17:48 |
OXPHOS | http://stackoverflow.com/questions/42719117/how-to-exercise-tensorflow-xla-aot-support-in-tensorflows-distribution | 17:48 |
CaBa | .oO(and i don't want to read any further ideas for improvements! :P) | 17:48 |
OXPHOS | wiking took my old laptop quite a while to run bazel | 17:49 |
@wiking | :))))))))))) | 17:49 |
@wiking | yeah bazel is a bitch | 17:49 |
@wiking | it sucks up like couple of gigs | 17:49 |
@wiking | as a starter | 17:49 |
@wiking | :) | 17:49 |
OXPHOS | indeed | 17:50 |
@wiking | but yeah | 17:50 |
@wiking | after the first time | 17:50 |
@wiking | things are smooth | 17:50 |
@wiking | :) | 17:50 |
OXPHOS | sounds | 17:50 |
OXPHOS | weird | 17:50 |
OXPHOS | lol | 17:50 |
@wiking | :D | 17:50 |
@wiking | i did not SAY that | 17:50 |
@wiking | just to be clear! | 17:50 |
@wiking | you did | 17:50 |
@wiking | (code of conduct! | 17:50 |
@wiking | ) | 17:51 |
OXPHOS | good thing without logs | 17:51 |
@wiking | :>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>. | 17:51 |
@wiking | anyhow | 17:52 |
@wiking | yeah | 17:53 |
@wiking | once you have the jit | 17:53 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 17:53 | |
@wiking | then only thing you need is to detect that | 17:53 |
@wiking | and the client.h | 17:53 |
@wiking | after that | 17:53 |
@wiking | you should be golden | 17:53 |
rads | ..lurking | 17:53 |
OXPHOS | wiking: I think I got that part already (hopefully) | 17:54 |
OXPHOS | installation :))) | 17:54 |
OXPHOS | thx wiking. | 17:54 |
@wiking | cool | 17:54 |
@wiking | then just detect XLA | 17:54 |
@wiking | no need for tf | 17:54 |
OXPHOS | sure | 17:54 |
@wiking | and start just throwing 1 by 1 | 17:55 |
@wiking | some simple operations on the jit | 17:55 |
@wiking | see if it works at all | 17:55 |
@wiking | i mean the shitty part is going to be the data munging | 17:55 |
@wiking | i believe | 17:55 |
@lambday | omigod... so much love.. | 17:56 |
@lambday | love, love everywhere | 17:56 |
@lambday | did I miss the *grabs popcorn* moments? | 17:57 |
@wiking | OXPHOS, https://github.com/tensorflow/tensorflow/blob/master/tensorflow/compiler/xla/array2d.h | 17:57 |
@HeikoS | CaBa: will check your prs soon | 17:57 |
@wiking | OXPHOS, i guess you could wrap sgmatrix T* into std::vector | 17:57 |
@wiking | or? | 17:57 |
@wiking | &vector[0] will give you the array | 17:58 |
CaBa | HeikoS: as long as you don't get any ideas... :D | 17:58 |
@wiking | lambday, can you do the wrapping actually? | 17:58 |
@lambday | wiking: what wrapping? | 17:58 |
@lambday | sorry I wasn't following | 17:58 |
@HeikoS | also all data structures in SG should be just wrapped std::vectors :) | 17:58 |
@lambday | HeikoS: huh? when did we decide on that? | 17:58 |
@HeikoS | lambday: didnt | 17:59 |
@HeikoS | but I think it is a good idea | 17:59 |
@wiking | lambday, say i have a void* array | 17:59 |
@lambday | oh I thought I missed some good discussion (like today's.. treat to read) | 17:59 |
@wiking | lambday, and wanna somehow wrap that into a std::vector | 17:59 |
@lambday | wiking: type info? | 17:59 |
@wiking | what ever | 17:59 |
@wiking | doulble* | 17:59 |
@wiking | and i wanna just pass in the memory chunk | 18:00 |
@wiking | to std::vector | 18:00 |
@wiking | and have it wrapped | 18:00 |
@wiking | but not copied | 18:00 |
@lambday | okay.. so I'd std::move it | 18:00 |
@lambday | pass the ownership? | 18:00 |
@wiking | coz vector.push_Back | 18:00 |
@lambday | I mean, should std::vector be responsible for free-ing it as well? | 18:00 |
@wiking | is not really a nice thing when the array lenght is 1k++ | 18:00 |
@lambday | or that memory chunk is referenced somewhere else | 18:01 |
@wiking | nono | 18:01 |
@wiking | just wrapping | 18:01 |
@lambday | okay | 18:01 |
@wiking | no need to do allocation/deallocation managment | 18:01 |
@lambday | wiking: I see | 18:01 |
@lambday | can be done | 18:01 |
@lambday | something like reference_wrapper | 18:02 |
@lambday | would that memory be modified? | 18:02 |
@lambday | wiking: | 18:02 |
@wiking | yep | 18:02 |
@lambday | okay | 18:02 |
@sukey | New Commit "Really proper way to handle W in OCAS SVM" to shogun-toolbox/shogun by lisitsyn: https://github.com/shogun-toolbox/shogun/commit/48e7ed44795a63a3e2c9a4bc12acf95641d4ac3f | 18:02 |
@sukey | Pull Request #3674 "Use weight vector *only* through getters/setters" synchronized by lisitsyn - https://github.com/shogun-toolbox/shogun/pull/3674 | 18:02 |
@lambday | wiking: so, shall I write a wrapper class.. or... ? | 18:03 |
@wiking | nono | 18:03 |
@wiking | just a question | 18:03 |
@wiking | how would it go | 18:03 |
@wiking | say | 18:03 |
@wiking | SGMatrix.matrix | 18:03 |
@wiking | to be wrapped around by std::vector | 18:03 |
@lambday | wiking: doable | 18:04 |
@lambday | I think | 18:04 |
@lambday | wiking: so, my question is : where would that method go? cause this can apply to any memory chunk.. doesn't need to be inside SGMatrix.. thus, a util class method or simply a method within shogun namespace? | 18:07 |
@wiking | lambday, this is XLA <-> SGMatrix | 18:07 |
@wiking | :) | 18:07 |
@lambday | oh | 18:07 |
@lambday | so XLA needs it to be std::vector, eh | 18:08 |
@lambday | cool | 18:08 |
@wiking | yes | 18:08 |
@lambday | I say, let's write a separate method.. inside SGMatrix --> XLA map, we call it | 18:09 |
-!- magicwund [~goksinen@50.75.193.138] has joined #shogun | 18:09 | |
@wiking | i mean we can put it for now | 18:13 |
@wiking | as a static | 18:13 |
@wiking | into anywhere | 18:13 |
@wiking | :) | 18:13 |
-!- magicwund [~goksinen@50.75.193.138] has quit [Ping timeout: 260 seconds] | 18:13 | |
@wiking | as OXPHOS just works on this as PoC | 18:18 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 18:20 | |
@wiking | lisitsyn, товарищ | 18:22 |
@wiking | is the microservice scalable thing ready for deployment ? :) | 18:23 |
@wiking | we need 10 of those! | 18:23 |
-!- lambday [6a33177c@gateway/web/freenode/ip.106.51.23.124] has quit [Ping timeout: 260 seconds] | 18:25 | |
-!- Killin [~Killin-A1@129.7.0.237] has joined #shogun | 18:37 | |
-!- Killin [~Killin-A1@129.7.0.237] has quit [Ping timeout: 260 seconds] | 18:56 | |
-!- rads [3a92880d@gateway/web/freenode/ip.58.146.136.13] has quit [Quit: Page closed] | 19:07 | |
-!- mikeling [uid89706@gateway/web/irccloud.com/x-rrtimeiwrndkxqdo] has quit [Quit: Connection closed for inactivity] | 19:40 | |
-!- HeikoS [~heiko@untrust-out.swc.ucl.ac.uk] has quit [Ping timeout: 264 seconds] | 19:41 | |
-!- ikazu [a027de6e@gateway/web/freenode/ip.160.39.222.110] has quit [Quit: Page closed] | 19:41 | |
-!- abhinav_ [~abhinav@103.211.53.10] has joined #shogun | 19:51 | |
-!- abhinav_ [~abhinav@103.211.53.10] has quit [Quit: Leaving] | 20:07 | |
-!- travis-ci [~travis-ci@ec2-54-196-243-107.compute-1.amazonaws.com] has joined #shogun | 20:36 | |
travis-ci | it'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/213920981 | 20:36 |
-!- travis-ci [~travis-ci@ec2-54-196-243-107.compute-1.amazonaws.com] has left #shogun [] | 20:36 | |
-!- Eon [6715530a@gateway/web/freenode/ip.103.21.83.10] has joined #shogun | 20:44 | |
-!- Eon is now known as Guest87555 | 20:45 | |
-!- Guest87555 [6715530a@gateway/web/freenode/ip.103.21.83.10] has left #shogun [] | 20:45 | |
@sukey | New Commit "CRYPTABT-363 Do not use apply_one on intermediate values of w" to shogun-toolbox/shogun by lisitsyn: https://github.com/shogun-toolbox/shogun/commit/3a589db6f950e70c26cf5fdc80a49823d1b52492 | 20:53 |
@sukey | Pull Request #3674 "Use weight vector *only* through getters/setters" synchronized by lisitsyn - https://github.com/shogun-toolbox/shogun/pull/3674 | 20:53 |
lisitsyn | what? :D | 20:53 |
lisitsyn | I entered the ticket I work on at my job | 20:54 |
lisitsyn | hahaha | 20:54 |
@sukey | Pull Request #3674 "Use weight vector *only* through getters/setters" synchronized by lisitsyn - https://github.com/shogun-toolbox/shogun/pull/3674 | 20:54 |
@sukey | New Commit "Do not use apply_one on intermediate values of w" to shogun-toolbox/shogun by lisitsyn: https://github.com/shogun-toolbox/shogun/commit/70362c59fa5e338ca12245eeec206ace623d0292 | 20:54 |
-!- travis-ci [~travis-ci@ec2-54-198-124-107.compute-1.amazonaws.com] has joined #shogun | 20:56 | |
travis-ci | it'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/213981614 | 20:56 |
-!- travis-ci [~travis-ci@ec2-54-198-124-107.compute-1.amazonaws.com] has left #shogun [] | 20:56 | |
CaBa | lisitsyn: lol | 21:13 |
-!- tctara_ [~quassel@128.199.61.169] has quit [Ping timeout: 240 seconds] | 21:13 | |
-!- tctara_ [~quassel@128.199.61.169] has joined #shogun | 21:17 | |
lisitsyn | hah that was ridiculous | 21:21 |
-!- rads [3a92880d@gateway/web/freenode/ip.58.146.136.13] has joined #shogun | 21:24 | |
rads | hey guys! i have a question | 21:25 |
rads | is there an issue compiling shogun manually on a mac with anaconda3? somehow the python.h file is not being seen by shogun | 21:26 |
CaBa | never used anaconda | 21:37 |
CaBa | so you can't compile in the first place i take it? | 21:38 |
CaBa | the only problem i had with building sg incl py bindings on my mac was related to additional py installs other than the system py | 21:39 |
rads | the problem comes while compiling with the python interface | 21:39 |
CaBa | rads: so how exactly do you compile sg? | 21:43 |
rads | kinda did this cmake -DCMAKE_INSTALL_PREFIX=path/to/shogun/install/dir -DPYTHON_INCLUDE_DIR=path/to/anaconda/include/python2.7/ -DPYTHON_LIBRARY=path/to/anaconda/lib/libpython2.7.so -DPYTHON_EXECUTABLE=path/to/anaconda/bin/python -DPythonModular=On .. | 21:44 |
rads | to link to the anaconda distr | 21:45 |
CaBa | hm i have no experience with anaconda. maybe they have some virtual-env like mode where they configure your environment such that everything python related points to their distribution? | 21:47 |
-!- travis-ci [~travis-ci@ec2-54-159-45-232.compute-1.amazonaws.com] has joined #shogun | 21:49 | |
travis-ci | it'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/213981900 | 21:49 |
-!- travis-ci [~travis-ci@ec2-54-159-45-232.compute-1.amazonaws.com] has left #shogun [] | 21:49 | |
CaBa | there seems to be something like conda envs... maybe try to install shogun from one of those? | 21:49 |
rads | isn't that helpful when you need different python versions on one system? | 21:55 |
rads | as in, in anaconda itself | 21:56 |
lisitsyn | rads: I just used the same thing with my anaconda | 21:59 |
rads | which one?? | 22:01 |
lisitsyn | rads: well -DPYTHON_LIBRARY and so on | 22:01 |
rads | on mac? hmm i've been trying for some time now | 22:02 |
lisitsyn | rads: ah yeh it failed | 22:03 |
rads | how did you make it work? | 22:04 |
lisitsyn | rads: no I didn't | 22:05 |
rads | ok so what to do now? :P | 22:05 |
lisitsyn | rads: experimenting here :) | 22:10 |
rads | :P | 22:12 |
rads | it only works without the python binding | 22:13 |
lisitsyn | rads: no it worked | 22:14 |
lisitsyn | rads: show the options :P | 22:14 |
lisitsyn | -DPYTHON_INCLUDE_DIR=path/to/anaconda/include/python2.7/ | 22:14 |
lisitsyn | this one is wrong in your case | 22:15 |
lisitsyn | are you sure about that? | 22:15 |
rads | no that's not the exact thing | 22:15 |
rads | wait a sec | 22:15 |
lisitsyn | I mean the path itself that you used, does it contain the Python.h | 22:15 |
lisitsyn | yeah i get its a placeholder | 22:15 |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 22:15 | |
lisitsyn | just list the directory you used to check it has Python.h | 22:15 |
rads | -DPYTHON_INCLUDE_DIR= anaconda3/include/python3.6m/ | 22:16 |
lisitsyn | ~/? | 22:17 |
rads | it worked for the libraries | 22:18 |
lisitsyn | rads: use absolute path | 22:18 |
lisitsyn | or relative to your home | 22:18 |
lisitsyn | I think that could be an issue | 22:18 |
lisitsyn | or the issue :) | 22:19 |
-!- rads_ [3a92880d@gateway/web/freenode/ip.58.146.136.13] has joined #shogun | 22:21 | |
rads_ | lemme try this from the start.. | 22:22 |
-!- sphegde [80e6c326@gateway/web/freenode/ip.128.230.195.38] has joined #shogun | 22:22 | |
-!- sphegde_ [80e6c326@gateway/web/freenode/ip.128.230.195.38] has joined #shogun | 22:22 | |
-!- rads [3a92880d@gateway/web/freenode/ip.58.146.136.13] has quit [Ping timeout: 260 seconds] | 22:22 | |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has joined #shogun | 22:24 | |
rads_ | lisitsyn did you use source activate? | 22:25 |
lisitsyn | rads_: no | 22:25 |
lisitsyn | rads_: I think you've lost your connection? | 22:26 |
lisitsyn | I suggested to not use relative paths | 22:26 |
rads_ | yeah i did lose it..trying it again | 22:26 |
lisitsyn | rads_: use something like ~/anaconda3/include/python3.5m or so | 22:27 |
lisitsyn | rads_: because it could be in different directory when compliinng the stuff | 22:28 |
@sukey | New Commit "Use no apply_one in perceptron because it is not trained yet" to shogun-toolbox/shogun by lisitsyn: https://github.com/shogun-toolbox/shogun/commit/9913b33a81e5e280be3338b4eda0803f63e87357 | 22:30 |
@sukey | Pull Request #3674 "Use weight vector *only* through getters/setters" synchronized by lisitsyn - https://github.com/shogun-toolbox/shogun/pull/3674 | 22:30 |
-!- sphegde_ [80e6c326@gateway/web/freenode/ip.128.230.195.38] has quit [Quit: Page closed] | 22:47 | |
-!- micmn [uid216702@gateway/web/irccloud.com/x-jwnhkzyvxkgkmtsm] has quit [Quit: Connection closed for inactivity] | 22:48 | |
rads_ | daaang | 23:01 |
@sukey | New Commit "Use weight vector *only* through getters/setters" to shogun-toolbox/shogun by lisitsyn: https://github.com/shogun-toolbox/shogun/commit/9053a8f01c9f0fbaaf1df4136bc10532e2cda922 | 23:06 |
@sukey | Pull Request #3674 "Use weight vector *only* through getters/setters" synchronized by lisitsyn - https://github.com/shogun-toolbox/shogun/pull/3674 | 23:06 |
-!- sphegde [80e6c326@gateway/web/freenode/ip.128.230.195.38] has quit [Ping timeout: 260 seconds] | 23:20 | |
-!- magicwund [~goksinen@rrcs-50-75-193-138.nyc.biz.rr.com] has quit [Remote host closed the connection] | 23:35 | |
-!- rads_ [3a92880d@gateway/web/freenode/ip.58.146.136.13] has quit [Quit: Page closed] | 23:42 | |
--- Log closed Thu Mar 23 00:00:16 2017 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!