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/
darthgera has quit [Remote host closed the connection]
< Nakul[m]> > : I have a report callback almost finished, but that could be part of another release.
< Nakul[m]> I think this is also finished https://github.com/mlpack/mlpack/pull/2075
< Nakul[m]> @zoq have a look
< KhizirSiddiquiGi> Nakul: CallbackTest doesn't pass, try to fix it using the Azure builld logs.
darthgera has joined #mlpack
< darthgera> warning: libmkl_rt.so, needed by /usr/lib/x86_64-linux-gnu/libarmadillo.so, not found (try using -rpath or -rpath-link)
< darthgera> make has failed and its showing this
< jenkins-mlpack2> Project docker mlpack nightly build build #583: STILL UNSTABLE in 2 hr 57 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/583/
< KhizirSiddiquiGi> darthgera: Try running `$ locate libmkl_rt.so` on terminal. It should show some locations of libmkl_rt.so.
< darthgera> yeah it showed in anaconda and i tried linking it to it
< darthgera> but so far no progress. i did export LD_LIBRARY_PATH=/home/gera/anaconda3/lib/libmkl_rt.so
< darthgera> and did cmake and make
< Nakul[m]> Khizir Siddiqui (Gitter): but when I run ctest -T it passes all test cases
xiaohong has joined #mlpack
< zoq> darthgera: So you solved the armadillo build issue (including libmkl), you could also build armadillo against openblas.
< zoq> Nakul[m]: Did you build mlpack with -DDEBUG=ON, as I mentioned before, if you don't some checks are disabled.
< darthgera> yes the armadilo issue was resolved but it fails at 57% with this error
< darthgera> still unable to link and getting tons of warnings
< zoq> darthgera: can you post the error message?
< zoq> darthgera: Right, but I thought you got the same warning in the armadilo build?
< darthgera> while building armadilo i didnt get any warnings. during make, initially libmkl come as warnings and then at 57% it fails because of it
< zoq> If I remember right, one of the error message mentioned setting the rpath, did you test that?
< darthgera> it was a warning, could you tell how to set the rpath please
< zoq> -DCMAKE_EXE_LINKER_FLAGS="-rpath -Wl,/usr/local/lib", whatever the path is
< darthgera> infront of cmake?
< zoq> right
xiaohong_ has joined #mlpack
xiaohong has quit [Ping timeout: 272 seconds]
< darthgera> its not working
< darthgera> unable to run cmake
< KhizirSiddiquiGi> darthgera: Try `pip install mkl`.
< zoq> what command line did you use?
< darthgera> cmake -DCMAKE_EXE_LINKER_FLAGS="-rpath -Wl,/usr/home/gera/anaconda3/lib/libmkl_rt.so" ..
< zoq> darthgera: it should be a path not the file: cmake -DCMAKE_EXE_LINKER_FLAGS="-rpath -Wl,/usr/home/gera/anaconda3/lib/ -lmkl_intel_ilp64 -lmkl_core -lmkl_intel_thread -lpthread" ..
< rcurtin> darthgera: LD_LIBRARY_PATH should be a directory, not the name of the file, so maybe /home/gera/anaconda3/lib/ is fine for that
< rcurtin> also, are you sure it should be /usr/home/? I haven't seen a home directory under /usr since... some Solaris variant or ancient UNIX derivative...?
< rcurtin> I guess Solaris and SunOS used /export/home not /usr/home
< rcurtin> hmm... it appears I accidentally released ensmallen 2.11.2 dated tomorrow
< rcurtin> but, it's 2:30am on the 16th in Japan, so, I guess it's okay :)
< KhizirSiddiquiGi> rcurtin: :laughing:
< rcurtin> I should have written in the release notes... "this release is from the future!"
< KhizirSiddiquiGi> and then `import ensmallen from future`
< rcurtin> :)
< KhizirSiddiquiGi> :)
< jeffin143[m]> 😂
ImQ009 has joined #mlpack
< hemal[m]> Hello @zoq, @rcurtin. I would like to work on Graph Convolution Networks as a part of GSOC 2020. It could be considered as a part of “Essential Deep Learning Modules” or independent. My preliminary research says that it should be a part of MLpack.
< hemal[m]> 1. Has there been any discussion on this topic?
< hemal[m]> Let me know what do you think, thanks!
< rcurtin> if we want to provide graph CNNs, we need to ensure that (a) the user interface is consistent with the rest of mlpack and (b) the way that we represent the graphs that get passed as input is reasonable
< rcurtin> given that everything in Armadillo is expressed as matrices, this could be tricky, but if you think about it and come up with something consistent, I think it could be a good direction to go
< rcurtin> now, do be aware that we haven't even applied to GSoC 2020 yet, and we have no idea if we'll be accepted
darthgera has quit [Remote host closed the connection]
< hemal[m]> Yes, I am going through the paper, and I feel that I could come up with something consistent.
< hemal[m]> Ok, I'll start working on it anyways. :)
< rcurtin> sounds good
cccdnl has joined #mlpack
cccdnl has quit [Remote host closed the connection]
darthgera has joined #mlpack
< darthgera> cmake command still throwing errors after fixing path
< darthgera> is there a way where it can automatically install to usr/local/lib
< rcurtin> darthgera: what have you tried to fix it? have you looked through the CMake documentation to see how to specify install paths?
< rcurtin> just as a note, it's very hard for people to help when the input we get is "the command did not work" and not precise error messages
< rcurtin> :)
< darthgera> sorry rcurtin. right so you suggested earlier that the path i was providing to cmake flag was incorrect which i fixed it
< darthgera> i also tried by changing export LD_LIBRARY_PATH="/home/anaconda3/lib/:$LD_LIBRARY_PATH"
< darthgera> this in turn has now started throwing /usr/bin/cmake: /home/gera/anaconda3/lib/libcurl.so.4: no version information available (required by /usr/bin/cmake)
< darthgera> however, libmkl_rt.so warnings are no longer coming(which is a relief)
gtank___ has quit [Read error: Connection reset by peer]
gtank___ has joined #mlpack
gtank___ has quit [Read error: Connection reset by peer]
< zoq> darthgera: Maybe it is easier to not build against mkl?
< darthgera> how do i do that?
< darthgera> cmake flag?
gtank___ has joined #mlpack
< zoq> darthgera: https://gist.github.com/BERENZ/ff274ebbf00ee111c708 should be helpful.
< darthgera> thanks, will check it out. rn its compiling with that warning
< Nakul[m]> > Nakul: Did you build mlpack with -DDEBUG=ON, as I mentioned before, if you don't some checks are disabled.
< Nakul[m]> yeah and tried once again ctest -T here is the result https://pastebin.com/CHxEr4Pf but can't understand why it is failing
< darthgera> i did with -ddebug=on earlier but didnt work
ImQ009 has quit [Quit: Leaving]
< Nakul[m]> * > Nakul: Did you build mlpack with -DDEBUG=ON, as I mentioned before, if you don't some checks are disabled.
< Nakul[m]> yeah and tried once again ctest -T here is the result https://pastebin.com/CHxEr4Pf but can't understand why it is failing as you can see at line number 35
SachaD has joined #mlpack
aleksandr has joined #mlpack
SachaD has quit [Remote host closed the connection]
aleksandr has quit [Client Quit]
SachaD has joined #mlpack
< SachaD> Hello. I want to learn mlpack. I tried fnn example. I realized that mlpack uses arma::mat matrices, that has type double. Can I use mlpack with matrices of ushort type and how? arma::Mat<ushort> does not working. I saw lessons on youtube and there was told that is faster and better to use 2 or 4 byte types. Double is 8 bytes.
< darthgera> do i need to do make install after make,seems to do the same thing
< SachaD> darthgera, make - compiles the source code to libraries. make install copies them to the right place. In mlpack, I don't know why, make install does not set executable attribute for libraries. You will need to do that manually. By default make install copy them to /usr/local/lib
< darthgera> sudo apt install method doesnt do that?
< SachaD> it does if you install .deb package. If you want to install the latest version of something, you, usually, need to compile it by yourself
darthgera has quit [Remote host closed the connection]
SachaD has quit [Quit: Konversation terminated!]