mlpack IRC logs, 2020-04-16

Logs for the day 2020-04-16 (starts at 0:00 UTC) are shown below.

April 2020
Sun
Mon
Tue
Wed
Thu
Fri
Sat
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
--- Log opened Thu Apr 16 00:00:15 2020
07:07 -!- ImQ009 [~ImQ009@unaffiliated/imq009] has joined #mlpack
09:28 -!- marko [6d2903d9@ip-109-41-3-217.web.vodafone.de] has joined #mlpack
09:28 < marko> hi
09:28 < marko> this throws
09:29 < marko> read access violation
09:29 < marko> arma::mat features(11, 100);
09:29 < marko> arma::rowvec targets(1, 100);
09:30 < marko> mlpack::regression::LARS glm (false, .5, .5);
09:30 < marko> std::vector<mlpack::regression::LARS> inventoryofModels;
09:30 < marko> inventoryofModels.emplace_back(glm);
09:30 < marko> inventoryofModels[0].Train(features, targets);
09:31 < marko> LARS object is placed into a vector
09:31 < marko> and than accessed in the vector
09:31 < marko> in order to apply "Train" method
09:31 < marko> and this throws an exception
09:32 < marko> W10, VS2017
09:32 < marko> the debug is indicating line 235 in LARS.cpp.
09:32 < marko> matGramActive(i, j) = (*matGram)(activeSet[i], activeSet[j]);
10:12 < marko> anyone over there?
10:12 < marko> :)
12:28 < Manav-KumarGitte> Does mlpack have rule base classifier like ripple method?
12:29 < marko> trees are the closest you can find to rule based
12:29 < marko> browse here
12:29 < marko> https://www.mlpack.org/doc/mlpack-3.3.0/cli_documentation.html#method-specific-tutorials
12:35 < Manav-KumarGitte> But I guess they don't have rule based classifier which applies fuzzy logic in them?
13:31 < marko> if you cannot find it in docs, its not there
13:31 < marko> ;)
13:34 < rcurtin> marko: looks like maybe LARS needs a correct copy constructor? you could try something like this as a workaround:
13:34 < rcurtin> basically the idea would be to use std::vector<mlpack::regression::LARS*> instead (i.e. a vector of pointers) so that you can avoid the copy constructor
13:34 < rcurtin> it's not a great solution, but it's at least a workaround
13:34 < rcurtin> if you want to report this as a bug on github that would be great :)
13:43 < marko> let me try if that works
13:43 < rcurtin> you'll want to allocate each LARS object with `new` and make sure to `delete` it when the program is over :)
13:43 < marko> rcurtin many thanks
13:43 < marko> lets try
13:43 < rcurtin> sure; sorry that the copy constructor is not correct to begin with :)
13:44 < marko> khm...
13:45 < marko> * seems not to help
13:46 < marko> can you please clarify the second approach
13:46 < rcurtin> sure, hang on
13:47 < rcurtin> https://pastebin.com/22cWMyf9
13:47 < rcurtin> (then don't forget to call `delete inventoryofModels[0]` when you're done with everything)
13:48 < rcurtin> another approach would be to call Train() before emplace_back(), but I don't know if that works with your situation
13:50 < marko> is push_back mandatory)
13:50 < marko> ?
13:51 < rcurtin> hmmm, maybe emplace_back() will work with the pointer, but basically I want to avoid calling the copy constructor of LARS
13:51 < rcurtin> give it a shot, see what happens :)
13:53 < marko> i have placed the error in the github
13:53 < marko> https://github.com/mlpack/mlpack/issues/2372
13:53 < marko> there must be a workaround
13:53 < marko> Train does not work
13:53 < marko> Predict does work, that is funny
13:57 < marko> rcurtin
13:57 < marko> 'initializing': cannot convert from 'mlpack::regression::LARS' to '_Objty'
13:57 -!- conrad41 [674de893@103.77.232.147] has joined #mlpack
13:57 < rcurtin> does it say what _Objty is?
13:58 < rcurtin> also, what line gives that error?
13:58 < rcurtin> I just guessed at my workaround code, I didn't try it. so maybe some extra debugging is needed :)
13:59 < marko> thanks anyway
13:59 < marko> :D
13:59 < marko> here is the github error
13:59 < marko> https://github.com/mlpack/mlpack/issues/2372
14:00 < marko> rcurtin so we may continue over there
14:00 < marko> thanks
14:00 -!- marko [6d2903d9@ip-109-41-3-217.web.vodafone.de] has left #mlpack []
14:02 < rcurtin> marko: sounds good---I'll see if I can find some time to look later today
14:22 -!- Yashwants19 [dfbd8940@223.189.137.64] has joined #mlpack
14:27 < Yashwants19> Hi rcurtin, I saw something here https://www.mlpack.org/doc/mlpack-3.3.0/julia_documentation.html#sparse_coding. I think so we are not providing these options (`copy_all_inputs=false, help=false, info="", version=false`) with julia.
14:27 < Yashwants19> Or may be I misunderstood something here.
14:28 -!- Yashwants19 [dfbd8940@223.189.137.64] has quit [Remote host closed the connection]
14:29 < rcurtin> Yashwants19: interesting, `verbose` is valid, but the other three aren't; I'll look into why those are being printed
14:29 < rcurtin> the `info`, `help`, and `copy_all_inputs` inputs aren't actually a part of the Julia binding in the .jl, so it must be a printing bug
14:31 -!- Yashwants19 [dfbd8940@223.189.137.64] has joined #mlpack
14:32 -!- conrad41 [674de893@103.77.232.147] has left #mlpack []
14:32 < Yashwants19> I think so we are allowing all persistent there.
14:33 < Yashwants19> We must allow `verbose` only.
14:33 < Yashwants19> *persistent options
14:41 < Yashwants19> I think so here https://github.com/mlpack/mlpack/blob/3ce835cbabd665d40ad11bbb1f841118c7a16702/src/mlpack/bindings/julia/print_doc_functions_impl.hpp#L478
14:45 -!- Yashwants19 [dfbd8940@223.189.137.64] has quit [Ping timeout: 240 seconds]
14:52 < rcurtin> Yashwants19: if you want to beat me to it and open a PR feel free :) I'm pretty busy today so it might be a bit
14:52 < rcurtin> I think you're right about the solution
15:00 -!- Yashwants19 [dfbd8940@223.189.137.64] has joined #mlpack
15:03 < Yashwants19> Great I will open a PR soon :)
15:04 -!- Yashwants19 [dfbd8940@223.189.137.64] has quit [Remote host closed the connection]
17:28 -!- PrinceGuptaGitte [gitterprin@gateway/shell/matrix.org/x-rxqsvvhlcxrbxijv] has joined #mlpack
17:28 * PrinceGuptaGitte sent a long message: < https://matrix.org/_matrix/media/r0/download/matrix.org/ZXCvbIaJMdDNTOxAVnvWvxbI >
17:28 < rcurtin> PrinceGuptaGitte: looks like it found the headers for Armadillo, but not libarmadillo.so; maybe you need to specify `-DARMADILLO_LIBRARY=/path/to/libarmadillo.so` in your CMake configuration?
17:31 < PrinceGuptaGitte> Thanks that worked. Now I'm getting "Unable to find the requested Boost libraries.", I suppose it's the same problem. But why could it be happening? I've never had to explicitly specify path to all the libraries during `cmake ../`
17:32 < rcurtin> did you clean your build directory after upgrading your system? you might want to do that
17:32 < rcurtin> i.e. either remove the entire build directory and make a new one, or, `rm CMakeCache.txt` will probably work too
17:33 < PrinceGuptaGitte> I did clean my build directory when I was getting this error.
17:34 < rcurtin> if that's happening from a clean build directory, I'm not sure what the issue is; I guess you can work around it by manually specifying the location of the Boost libraries
17:34 < rcurtin> if you find something wrong in the CMake configuration, feel free to open an issue, but I don't think that anything's wrong... I don't think we've changed how dependencies are found recently?
17:35 < PrinceGuptaGitte> Yeah, I guess I'll have to stay with that for some time.
17:45 < PrinceGuptaGitte> I just saw my PR #2184 went stale. What should I do about this?
17:47 < rcurtin> you can comment on it or push commits and it'll become un-stale; I see that there are merge conflicts, so if you can fix those it would be great
17:47 < rcurtin> personally I haven't had any time to review it
18:02 < rcurtin> I'm trying to figure out what's going on with the video meeting... is anyone else successfully joined?
18:02 < rcurtin> oh, sorry, stupidity on my end
18:54 -!- ImQ009 [~ImQ009@unaffiliated/imq009] has quit [Quit: Leaving]
19:01 < rcurtin> ahhh, those ensmallen stickers I just ordered... delayed: "These items will ship when normal service resumes." so who knows when they'll get here :)
19:02 < zoq> I wonder if that is the same here in Germany.
19:06 < zoq> Intersting, I think they got more expensive since the last time I checked, they want 105 Euro for 1k.
19:07 < jeffin143[m]> Anybody from Russia ?
19:07 < rcurtin> yeah, I got 1k stickers for $188 USD last time, this time it was $210
19:08 < zoq> rcurtin: for the circle sticker, 25mm x 25mm?
19:09 < zoq> Maybe I should order the batch? $210 is even more expensive.
19:09 < rcurtin> 2" x 2" (so ~50mm x 50mm)
19:09 < rcurtin> the mlpack stickers that I ordered in the past are that size, so I figured I'd go with the same size
19:09 < zoq> Ahh, that is 170 Euro, here
19:10 < rcurtin> it scales sublinearly with squared mm :)
19:11 < zoq> I don't mind to order some, but if we ask NumFOCUS to reimburse us, it doesn't matter.
19:11 < rcurtin> yeah, I'll get in touch with NumFOCUS and see what they think
19:12 < rcurtin> I figured I'd mail a bunch of the ones that I got to you, so that we have two locations we can send from :)
19:12 < zoq> rcurtin: Happy to have some here and mail them out :)
19:15 < rcurtin> :)
19:26 < sreenik[m]> It was a very cool discussion today, thanks Shikhar and Ryan (both :D)
19:31 < rcurtin> :)
19:36 < zoq> The sessions terminates after some time?
19:36 < rcurtin> the video chat? it should actually just be an always-open zoom room that anyone can use at any time
19:37 < sreenik[m]> It terminates after the host leaves I guess (i.e., if the host doesn't hand over before leaving)
19:37 < rcurtin> really?? sorry about that, I thought I had it configured differently
19:37 < sreenik[m]> No let me see if I can join now
19:38 < sreenik[m]> We hanged out for around half-an hour after you left :)
19:38 < rcurtin> ahhh, ok, that's good, it would be bad if it terminated immediately
19:39 < jeffin143[m]> I remember Omar was the host
19:39 < jeffin143[m]> After rcurtin lefy
19:39 < sreenik[m]> Hehe
19:39 < jeffin143[m]> Left*
19:39 < sreenik[m]> I checked, yes, we can join anytime
19:39 < jeffin143[m]> So now the power is with shrit to terminate the room
19:40 < jeffin143[m]> May it be good , two guys working on a pr, to pair programming sharing screen on always opened zoom room
19:40 < rcurtin> yeah, the zoom room would be perfect for things like that, feel free to use it that way :)
19:40 < jeffin143[m]> 🤗
19:41 < sreenik[m]> I think it'll be very effective during the gsoc
20:13 -!- yernar[m] [yernarmatr@gateway/shell/matrix.org/x-yjtivlxytpfuqwxr] has joined #mlpack
--- Log closed Fri Apr 17 00:00:16 2020