ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
gotadachi has quit [Quit: Leaving...]
< zoq> rcurtin[m]: Does https://gist.github.com/zoq/8d65e3caf0a58ef9772e8466caeaf3cc look familiar to you?
< rcurtin[m]> yes, it does, but I can't remember exactly why or when that happens... can you tell me about the CMake configuration you used?
< zoq> rcurtin[m]: It's a conda env, CMakeCache file - https://gist.github.com/zoq/ed37521cd30884e028d397b2ddb21740 if I build without the python bindings it's fine.
< rcurtin[m]> any chance you could also paste the generated `setup.py` in `build/src/mlpack/bindings/python/`?
< zoq> Looks like I'm not the first one - https://github.com/mlpack/mlpack/issues/1567
< rcurtin[m]> also, what happens if you specify `-DUSE_OPENMP=ON`?
< zoq> For some reason it can't find libpthread.so.
< rcurtin[m]> I think the actual error is right here in the gcc command: `-DBINDING_TYPE=BINDING_TYPE_PYX -std=c++11 -Wall -Wextra `
< zoq> That's another issue I have to figure out.
< rcurtin[m]> hm, it didn't paste right, but it appears as two spaces in the original output
< rcurtin[m]> so, another way to test here would just be to remove the `'${OpenMP_CXX_FLAGS}'` entry from that list (it's part of an if/else statement, you'd need to do it for the else case too)
< rcurtin[m]> if you want to give that a shot and see if it works, based on the result, I can put together a cleaner patch 👍️
< zoq> Yes that worked, thanks!
< rcurtin[m]> yep, I'm pretty sure that setup.py.in fix I suggested will work then
< rcurtin[m]> ah perfect
< rcurtin[m]> I'll write up a patch tomorrow 👍
< jonpsy[m]> Is there a hard limit on how long a ```_impl.hpp``` can be? That would explain a lot of errors I have been getting in AppVeyor. Thanks
< RishabhGargGitte> Hello CssTeja, as your error says, cmake is not able to find `cereal`. `Cereal` is a dependency of mlpack.
< RishabhGargGitte> So, if you haven't yet installed it, then you can simply head over to https://uscilab.github.io/cereal/quickstart.html and download the v1.3.0 zip file. Then simply extract it to whatever location you like. Then if let say, you extracted it in `D:/`. If you will see the contents of that folder, then you will find a folder called `include` which contains a folder called `cereal` inside it. Cereal is a header only C++
< RishabhGargGitte> library and including all the contents of that cereal folder will be enough. So, you can simply add the option `-DCEREAL_INCLUDE_DIR=D:/path/to/include` to the `cmake ../` command and that should work.
< jonpsy[m]> <jonpsy[m] "Is there a hard limit on how lon"> Around 350-450
< RishabhGargGitte> @zoq @rcurtin , many of the new developers of mlpack face this issue of installing cereal. Shall we put a guide about how to configure cereal for mlpack or how about creating an FAQ page where we can put answers to many more such beginner FAQs ? This would really help the newcomers.
gotadachi has joined #mlpack
< CssTejaGitter[m]> @RishabhGarg108 hey, i did download cereal 1.3.0 and added the option to cmake command to point to the include folder of cereal , but the error says found cereal but the version is 0.0.0, required atleast 1.1.2, even if the downloaded version is 1.3.0.
ImQ009 has joined #mlpack
hasan1101 has joined #mlpack
< MayankRaj[m]> Hey, can anyone help me with #2873 idk why this pr is failing.
< hasan1101> hello. i'm looking for gsoc projects. i want to work on improving mlpack's tree ensemble support. i have worked with decision trees, random forests and xgboost before. i can work on this. moreover, i also want to contribute to the example zoo (using jupyter notebooks).
hasan1101 has quit [Quit: Connection closed]
hasan1101 has joined #mlpack
hasan1101 has quit [Client Quit]
hasan1101 has joined #mlpack
hasan1101 has quit [Client Quit]
ImQ009 has quit [Read error: Connection reset by peer]
hasan1101 has joined #mlpack
hasan1101 has quit [Client Quit]
< rcurtin[m]> CssTeja (Gitter) can you tell us the command you used for CMake? what did you set `CEREAL_INCLUDE_DIR` to?
< RishabhGargGitte> Hello Ryan, I have opened a PR #2883 which attempts to add random split for numerical attributes in a decision tree. I don't want to rush you, you can take a look at it whenever you get time. But it could be great if you could provide feedback just on the test coverage. I have a few questions about it which I have mentioned on the PR.
lazaros65 has joined #mlpack
ImQ009 has joined #mlpack
lazaros65 has quit [Quit: Connection closed]
jayant has joined #mlpack
jayant has quit [Ping timeout: 240 seconds]
< rcurtin[m]> Rishabh Garg (Gitter): sure, I'll take a look when I'm able to 👍️
hasan1101 has joined #mlpack
hasan1101 has quit [Client Quit]
hasan1101 has joined #mlpack
hasan1101 has quit [Client Quit]
< jonpsy[m]> I have a question, inspired from NSGA-II code, instead of if/else block how about ```arma::uvec idxCondition = ( a < b); matrix % idxCondition```
< jonpsy[m]> nvm ill create an isuse
< zoq> jonpsy[m]: Depends on the context, it's not necessarily faster, and maybe not as easy to read.
< zoq> jonpsy[m]: A comment on the line in the code is a good idea.
< zoq> If you like you can time at as well.
< jonpsy[m]> i was implementing SBX crossover for SPEA-II and it has way too many if elses
< jonpsy[m]> using that tactic, i got rid of ```if``` block, **but**, there was a zero division (turns to nan)
< jonpsy[m]> bt its handled later in the algo, is it acceptable to have ```nan``` durin runtime of algo?
< zoq> If the function is called often, which might be the case here, I think it makes sense to do a simple benchmark to check the timings.
< zoq> If it's handled later, it's fine.
< jonpsy[m]> > If the function is called often, which might be the case here, I think it makes sense to do a simple benchmark to check the timings.
< jonpsy[m]> im not worried abt speed, more abt the readibility
< jonpsy[m]> would you mind if i create an issue? so that we're on the same page
< zoq> If you are not familiar with armadillo, or have a math background I don't think (a < b) % is super easy to read.
< zoq> We can comment on the PR itself I guess?
< jonpsy[m]> hmm...
< zoq> I don't think there is a general rule.
< jonpsy[m]> > We can comment on the PR itself I guess?
< jonpsy[m]> SPEA-II is currently WIP
ImQ009 has quit [Quit: Leaving]
< jonpsy[m]> on my lappy :)
< jonpsy[m]> * on my laptop :)
< zoq> You can also open a PR and mark it as WIP, but if you think an issue is easier that works as well, but make sure to provide some context.
< jonpsy[m]> cool, and one more thing.. is there a hard limit on length of code for ```_impl.hpp```
< jonpsy[m]> because my MOEA/D-DE suddenly started failing on windows Appveyor after it exceeded some 450 lines of code.
< zoq> There is no limit no.
< zoq> That is strange, I'll take a look into the issue later today.
< jonpsy[m]> is there a way to acess appveyor? its rlly strange..
< zoq> jonpsy[m]: You can SSH into the appveyor machine.
< jonpsy[m]> thats exactly what i need!
< jonpsy[m]> thanks for all this, see you tommorow, night!
izlacociredef has joined #mlpack