--- Log opened Fri Feb 16 00:00:04 2018 | ||
@wiking | lisitsyn, https://github.com/facebookresearch/TensorComprehensions jit is everywher :) | 09:04 |
---|---|---|
lisitsyn | wiking: ugly as shit! | 09:05 |
@wiking | yeah | 09:05 |
@wiking | i've looked into the code | 09:05 |
@wiking | it looks pretty bad api | 09:05 |
@wiking | The memory layout implied by TC is row-major (C-like). | 09:05 |
@wiking | wonder why | 09:05 |
lisitsyn | they use halide but halide is so freaking nice | 09:06 |
@wiking | :) | 09:06 |
@wiking | halide = llvm based? | 09:06 |
lisitsyn | Func blur_3x3(Func input) { | 09:06 |
lisitsyn | Func blur_x, blur_y; | 09:06 |
lisitsyn | Var x, y, xi, yi; | 09:06 |
@wiking | yep | 09:06 |
@wiking | just checking | 09:06 |
lisitsyn | much better huh | 09:07 |
@wiking | think we foudn a swig bug yesterday with Heiko :( | 09:07 |
@wiking | for ruby | 09:07 |
lisitsyn | seen some array stuff | 09:08 |
lisitsyn | vector/matrix/float something | 09:08 |
@wiking | yeah | 09:08 |
@wiking | that was ok | 09:08 |
@wiking | that was fixable | 09:08 |
@wiking | of course there was a bug in ruby typemap :) | 09:08 |
@wiking | but the problem is atm that when yo udo the dispatching with the generated swig | 09:09 |
@wiking | if you pass an object but the same api (polymorphism) accepts a pointer | 09:09 |
@wiking | there's a :kaboom: | 09:09 |
@wiking | William Fulton's reaction was: try latest swig from trunk :) | 09:09 |
@wiking | anyhow that's a blocker for ruby | 09:10 |
@wiking | but yeah i mean we should refactor ruby's typemaps | 09:11 |
@wiking | since | 09:11 |
@wiking | http://ruby-doc.org/stdlib-2.0.0/libdoc/matrix/rdoc/Matrix.html | 09:11 |
lisitsyn | aha | 09:11 |
@wiking | there's native matrix for ruby | 09:11 |
lisitsyn | wiking: fulton!!!!!!!!!! | 09:11 |
@wiking | since 2.0 | 09:11 |
@wiking | lisitsyn, heheh yeah swig guy :) | 09:11 |
lisitsyn | this guy | 09:11 |
lisitsyn | argh | 09:11 |
@wiking | heheh yeah he is not the easiest :) | 09:11 |
@wiking | lisitsyn, btw do you have an idea why we have taken jblas or ujmp as a jdk typemap? | 09:23 |
@wiking | i mean we could just have taken [] and [][] | 09:23 |
@wiking | that works in jdk | 09:23 |
lisitsyn | wiking: true | 09:23 |
lisitsyn | wiking: don't know | 09:23 |
lisitsyn | more cool? | 09:23 |
lisitsyn | :) | 09:23 |
@wiking | because both jblas and ujmp | 09:23 |
@wiking | are kind of not maintained anymore | 09:23 |
@wiking | :( | 09:23 |
@wiking | and i mean from [] and [][] | 09:24 |
@wiking | you could do *anything* in jvm | 09:24 |
lisitsyn | well nobody calculates stuff in java | 09:24 |
lisitsyn | :D | 09:24 |
@wiking | yeah but you would use as well | 09:24 |
@wiking | [][] and [] | 09:24 |
@wiking | and then maybe define wrappers | 09:24 |
@wiking | around | 09:24 |
lisitsyn | yeah that's true | 09:24 |
lisitsyn | although [][] is ineffective | 09:24 |
@wiking | yeah | 09:24 |
@wiking | i mean we can just say | 09:24 |
@wiking | [] | 09:24 |
@wiking | :) | 09:24 |
@wiking | and that it's column oriented | 09:25 |
@wiking | :p | 09:25 |
@wiking | i guess even less people would use us then from JVM | 09:25 |
@wiking | :DD | 09:25 |
lisitsyn | our matrices would be OBJECT oriented | 09:25 |
lisitsyn | because it is jvm | 09:25 |
@wiking | :D | 09:25 |
@wiking | i see you had woken up already :P | 09:25 |
@wiking | anyhow we coudl change dthat | 09:26 |
@wiking | ujmp Jul 30, 2015 | 09:26 |
@wiking | jblas May 12, 2015 | 09:26 |
@wiking | these are the release dates | 09:26 |
lisitsyn | uh | 09:27 |
lisitsyn | I guess we need to switch to either something else | 09:28 |
lisitsyn | or to [] | 09:28 |
@wiking | [] | 09:28 |
@wiking | as that lives forever | 09:29 |
@wiking | i dont see any other stuff | 09:29 |
@wiking | maybe commons math? | 09:29 |
lisitsyn | yeah apache something | 09:29 |
@wiking | i mean that at least has even sparse support as well | 09:29 |
@wiking | http://commons.apache.org/proper/commons-math/javadocs/api-3.6.1/org/apache/commons/math3/linear/SparseFieldMatrix.html | 09:29 |
lisitsyn | oh that sounds better | 09:30 |
lisitsyn | because commons is like everywhere | 09:30 |
@wiking | yeah | 09:32 |
@wiking | i mean it's ocmmons.math | 09:32 |
@wiking | bu tyeah | 09:32 |
@wiking | yaeh maybe that's the best | 09:32 |
@wiking | btw lisitsyn do you think it makes sense to have a mapping for vector as well? | 09:37 |
lisitsyn | wiking: not sure it makes sense everywhere | 09:37 |
lisitsyn | I mean why not assume vector is 1xd | 09:37 |
@wiking | i mean we can do the same as in ruby | 09:38 |
@wiking | there as i can see we both support | 09:38 |
@wiking | ruby array | 09:38 |
@wiking | and nmatrix | 09:39 |
@wiking | or what is that package | 09:39 |
@wiking | (this is at least when it's a typemap(in) | 09:39 |
@wiking | so we can support both [] and RealVector | 09:40 |
@wiking | and then drop ujmp and jblas | 09:41 |
-!- HeikoS [~heiko@bmi-airport-1.inf.ethz.ch] has joined #shogun | 10:00 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 10:00 | |
lisitsyn | HeikoS: welcome to zurich? | 10:25 |
lisitsyn | :) | 10:25 |
lisitsyn | wiking: HeikoS is visiting you? :) | 10:25 |
@HeikoS | lisitsyn: yes | 10:25 |
@HeikoS | I am leaving today | 10:25 |
@HeikoS | was here 3 days | 10:25 |
lisitsyn | oh I see | 10:25 |
-!- HeikoS [~heiko@bmi-airport-1.inf.ethz.ch] has quit [Ping timeout: 255 seconds] | 10:43 | |
@wiking | Trixis, here? | 11:24 |
@sukey | [https://github.com/shogun-toolbox/shogun] vigsterkr pushed 9 commits: | 16:46 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/3a3983cedf572fa06765281c22dbfb3082366e13 | 16:46 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/f474d37990f75a45f1d0c725ec8eb06f7aa3b558 | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/ab7f46630c1b4ed3d103d96d431c69e9d654e2e5 | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/b0acbde82a7ead6b5e000ff9eebf678f5e38609e | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/df6d3e4d9f8b953c701c45a958f26b5ecb486906 | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/f23601c0c7255d20a8b9429a13685ae60d072e55 | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/733289836748144ae437d70a30dda19a26245567 | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/c9dcbeb3d42fe349e847aed4a06884f4e6982a8b | 16:47 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/a65ae991155aead63e90828cc91d0867a1edb905 | 16:47 |
-!- HeikoS [~heiko@p4FED5061.dip0.t-ipconnect.de] has joined #shogun | 17:00 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 17:00 | |
@sukey | [https://github.com/shogun-toolbox/shogun] karlnapf pushed 2 commits: | 17:00 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/b1d3e0650e4e74a3d9613cf9fdf51152bebc6e82 | 17:00 |
@sukey | https://github.com/shogun-toolbox/shogun/commit/8f03a9dca3c8d620d8355720e4d3d0f4ca5c43e7 | 17:00 |
@sukey | [https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4177 | 17:01 |
@sukey | [https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4178 opened by karlnapf | 17:04 |
@sukey | [https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4178 karlnapf added label: "SWIG" | 17:04 |
@sukey | [https://github.com/shogun-toolbox/shogun] Issue https://github.com/shogun-toolbox/shogun/issues/4178 karlnapf added label: "good first issue" | 17:04 |
@HeikoS | lisitsyn: I think the build should now be green for "put" | 17:08 |
lisitsyn | HeikoS: wow you rock | 17:08 |
lisitsyn | :) | 17:08 |
@HeikoS | only thing that doesnt work is ruby, which I blacklisted for now | 17:08 |
@HeikoS | (only that example) | 17:08 |
@HeikoS | IF it is green, I think we shoujld merge it back | 17:09 |
@HeikoS | and then do another one for PUT | 17:09 |
@HeikoS | ehm | 17:09 |
@HeikoS | GET | 17:09 |
@HeikoS | lisitsyn: could you have a look at the diff of the feature branch to develop? | 17:09 |
lisitsyn | HeikoS: typedo? | 17:10 |
@HeikoS | I would do a PR for that, but wiking doesnt like doing PRs from feature branches, so not sure what is the best way | 17:10 |
@HeikoS | y | 17:10 |
lisitsyn | funny name | 17:10 |
lisitsyn | like tuxedo | 17:10 |
lisitsyn | but typedo | 17:10 |
lisitsyn | HeikoS: why not PR from feature branch? seems reasonable to me | 17:10 |
@HeikoS | it does to me too | 17:10 |
@HeikoS | travis works twice | 17:11 |
@HeikoS | but one gets the nice discussion webinterface for all the changes | 17:11 |
@HeikoS | Ill wait for travis and then open one | 17:11 |
@HeikoS | lisitsyn: or you wanna have a look now? | 17:11 |
lisitsyn | HeikoS: I checked the diff | 17:12 |
@HeikoS | ah ok | 17:12 |
lisitsyn | looks quite good | 17:12 |
@HeikoS | lets see whether it worked | 17:13 |
@HeikoS | and then Ill merge it | 17:13 |
@HeikoS | see you later! | 17:13 |
shogitter | (vikrant97) Hi, I am Vikrant Goyal and I am trying to setup the code in my PC but got some errors while doing "cmake .." in the build directory | 17:31 |
shogitter | (vikrant97) Could NOT find Doxygen (missing: DOXYGEN_EXECUTABLE) (Required is at least version "1.8.6") | 17:31 |
shogitter | (vikrant97) -- Found PythonInterp: /usr/bin/python (found version "2.7.12") | 17:31 |
shogitter | (vikrant97) CMake Error at src/shogun/CMakeLists.txt:49 (MESSAGE): | 17:31 |
shogitter | (vikrant97) [full message: https://gitter.im/shogun-toolbox/shogun?at=5a87075f17b16a2377566ebf] | 17:31 |
shogitter | (vikrant97) Can anyone please help? | 17:31 |
@wiking | HeikoS, doesn't like? :D | 17:32 |
@wiking | man what's the point of PR of a feature branch | 17:32 |
@wiking | when that was already tested by all the CIs | 17:32 |
@HeikoS | lisitsyn: ^ see :) | 17:33 |
@HeikoS | wiking: sure ;) | 17:33 |
@HeikoS | I guess only the nice summarising "changes" tab on github | 17:33 |
@HeikoS | as a central base for discussion | 17:33 |
@wiking | HeikoS, git squash? | 17:34 |
@wiking | as this is anyways one thing | 17:34 |
@wiking | and then you have one diff | 17:35 |
@HeikoS | I know | 17:35 |
@wiking | so i dont get it | 17:35 |
@wiking | or kill the CIs for the PR | 17:35 |
@HeikoS | yep I would do that | 17:35 |
@HeikoS | but I think we can merge without PR | 17:36 |
@HeikoS | once travis is fine | 17:36 |
@HeikoS | wiking: I blacklisted the example btw in ruby | 17:36 |
@wiking | HeikoS, mmm | 17:36 |
@wiking | could you not merge it plz | 17:36 |
@wiking | and give me couple of hours | 17:36 |
@HeikoS | sure | 17:37 |
@HeikoS | I dont want to merge it NOW :) | 17:37 |
@wiking | as at the moment i'm doing what will wants us to do | 17:37 |
@HeikoS | wiking: but would be good if we did that some time soon | 17:37 |
@HeikoS | sure sure | 17:37 |
Trixis | wiking: ye | 17:37 |
@HeikoS | wiking: no rush, just would like to get it in before I start working on "get" | 17:38 |
@wiking | you can have a branch from this to get branch | 17:38 |
@wiking | and then you'll just merge that | 17:38 |
shogitter | (vikrant97) can anyone please help? | 17:39 |
@wiking | HeikoS, you still have code in .cpp | 17:39 |
@wiking | regarding put | 17:39 |
@HeikoS | yep | 17:39 |
@HeikoS | wiking: think it should stay, as these methods will be extended soon | 17:40 |
@wiking | and it's still put_scalar | 17:40 |
@wiking | so i mean | 17:40 |
@HeikoS | put_scalar is fine in fact | 17:40 |
@wiking | this is not mergable | 17:40 |
@wiking | atm | 17:40 |
@wiking | and void put_vector_or_matrix | 17:40 |
@wiking | is still there | 17:40 |
@HeikoS | yep | 17:40 |
@HeikoS | would you mind fixing that? I couldnt | 17:41 |
@wiking | all? | 17:41 |
@HeikoS | the vector/matrix %extend | 17:41 |
@HeikoS | and if you can rename put_scalar to put, do it | 17:41 |
@HeikoS | oitherwise I think it doesnt matter | 17:41 |
@wiking | https://github.com/shogun-toolbox/shogun/blob/8f03a9dca3c8d620d8355720e4d3d0f4ca5c43e7/src/shogun/base/SGObject.h#L406 | 17:41 |
@wiking | thi sis not good | 17:41 |
@wiking | its not a good api | 17:42 |
@wiking | when you have in swig a general put | 17:42 |
@HeikoS | yep that as well | 17:42 |
@wiking | but in c++ is not good | 17:42 |
@wiking | because of whatever reason | 17:42 |
@wiking | but that language supports polymorphism | 17:42 |
@wiking | so i'm not so sure if there's a way to justify | 17:42 |
@wiking | why in c++ of all apis | 17:42 |
@HeikoS | ah yes sure | 17:42 |
@HeikoS | remove it | 17:42 |
@HeikoS | or I can do | 17:42 |
@wiking | put for scalar is put_scalar | 17:42 |
@wiking | but not in other langs | 17:42 |
@wiking | and yeah that type trait | 17:43 |
@wiking | is just not good | 17:43 |
@HeikoS | change it | 17:43 |
@wiking | as you dont want to have Some<CSGObject*> | 17:43 |
@HeikoS | I'm waiting for those fixes | 17:43 |
@wiking | just Some<CSGObject> | 17:43 |
@HeikoS | yep | 17:43 |
@wiking | so there's no need for typename std::remove_pointer<T>::type>::value | 17:43 |
@wiking | and that type trait since it's not swiged | 17:44 |
@wiking | it can be simply be | 17:44 |
@wiking | , std::enable_if_t<std::is_convertible<T, CSGObject>::value>* = nullptr> | 17:45 |
@HeikoS | wiking: would you mind fixing those in the branch? | 17:45 |
@HeikoS | I spend some time changing stuff but always broke the compilation or the interface | 17:45 |
@wiking | could you push these obvious fixes | 17:45 |
@wiking | that i have just told you | 17:45 |
@HeikoS | this version now at least works | 17:45 |
@wiking | but say that thing is buggy | 17:45 |
@HeikoS | if it is quick, pls just do it | 17:45 |
@wiking | :405 | 17:45 |
@wiking | :406 | 17:45 |
@wiking | so even if it 'worksforme' | 17:46 |
@wiking | it's just allowing something | 17:46 |
@wiking | that is prone to error | 17:46 |
@HeikoS | it would be amazing if you could just push all this quickly in there | 17:46 |
@wiking | ? | 17:47 |
@wiking | i've just given a review of the code | 17:47 |
@wiking | :) | 17:47 |
@wiking | a quick one | 17:47 |
@wiking | imo this is just not ready for merge | 17:47 |
@HeikoS | wiking: sure I get it, what I ask you, would you mind doing these 1 line changes? | 17:48 |
@HeikoS | would appreciate | 17:48 |
@wiking | yes | 17:48 |
@wiking | i would mind :) | 17:48 |
@wiking | as tehre's other stuff in this branch | 17:48 |
@wiking | that needs fixing | 17:48 |
@wiking | that are not so straightforward | 17:48 |
@wiking | say the put_scalar | 17:48 |
@wiking | so i rather look on that | 17:48 |
shogitter | (geektoni) ping wiking | 18:01 |
@wiking | geektoni pong | 18:01 |
-!- travis-ci [~travis-ci@ec2-54-167-157-129.compute-1.amazonaws.com] has joined #shogun | 18:01 | |
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/342399892 | 18:01 |
-!- travis-ci [~travis-ci@ec2-54-167-157-129.compute-1.amazonaws.com] has left #shogun [] | 18:01 | |
@wiking | Trixis, mmm i cannot recall why i pinged u | 18:01 |
@wiking | :( | 18:01 |
shogitter | (geektoni) wiking do you have a minute? ;) | 18:02 |
shogitter | (geektoni) I need to talk you about something | 18:02 |
@wiking | sure thing | 18:02 |
@wiking | on gitter? | 18:02 |
-!- witness [uid10044@gateway/web/irccloud.com/x-qzkhqtvckgzfmerp] has joined #shogun | 18:02 | |
shogitter | (geektoni) yes yes | 18:02 |
@wiking | ust a sec | 18:03 |
-!- travis-ci [~travis-ci@ec2-54-166-192-148.compute-1.amazonaws.com] has joined #shogun | 18:12 | |
travis-ci | it's Heiko Strathmann'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/342407176 | 18:12 |
-!- travis-ci [~travis-ci@ec2-54-166-192-148.compute-1.amazonaws.com] has left #shogun [] | 18:12 | |
-!- travis-ci [~travis-ci@ec2-54-166-192-148.compute-1.amazonaws.com] has joined #shogun | 18:17 | |
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/342399892 | 18:17 |
-!- travis-ci [~travis-ci@ec2-54-166-192-148.compute-1.amazonaws.com] has left #shogun [] | 18:17 | |
@HeikoS | vikrant97, pls read the installation readme and the error msg | 18:17 |
-!- HeikoS [~heiko@p4FED5061.dip0.t-ipconnect.de] has quit [Ping timeout: 248 seconds] | 18:30 | |
-!- travis-ci [~travis-ci@ec2-54-224-67-202.compute-1.amazonaws.com] has joined #shogun | 18:53 | |
travis-ci | it's Heiko Strathmann'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/342407176 | 18:53 |
-!- travis-ci [~travis-ci@ec2-54-224-67-202.compute-1.amazonaws.com] has left #shogun [] | 18:53 | |
-!- witness [uid10044@gateway/web/irccloud.com/x-qzkhqtvckgzfmerp] has quit [Quit: Connection closed for inactivity] | 20:21 | |
--- Log closed Sat Feb 17 00:00:05 2018 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!