verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
< zoq>
Looks like I have some problems to build the code on freebsd with the new TreeType API.
< zoq>
Is it okay to use the bound::HRectBound in #include <mlpack/core/tree/binary_space_tree/typedef.hpp> without template arguments?
< zoq>
And I guess we should add 'binary_space_tree/typedef.hpp', 'binary_space_tree.hpp' and 'cover_tree/typedef.hpp' to the CMakeLists.txt file.
< naywhayare>
zoq: it should be okay to use bound::HRectBound without template parameters, because the fourth template argument to BinarySpaceTree is a template template parameter
< naywhayare>
you're right that those files should be added to CMakeLists.txt... do you want to do that or should I?
< zoq>
I can commit the change
< naywhayare>
okay, thanks
< naywhayare>
what is the failure on FreeBSD? I can try to help figure it out
< naywhayare>
I'll try to compile on my FreeBSD box but the gcc I have there is 4.2, so it might be too old :)
< naywhayare>
(it will also take forever to compile, since it's a 400MHz sparc64 system)
< zoq>
I wonder why I get 'emst/dtb.hpp:79:38: error: default template argument for a template template parameter must be a class template' with a reference to class TreeType = tree::KDTree>
< zoq>
Which could be fixed by including '#include <mlpack/core/tree/binary_space_tree/typedef.hpp>' before '#include <mlpack/core/tree/binary_space_tree.hpp>' which makes no sense because binary_space_tree.hpp includes typedef.hpp.
< zoq>
Afterwards I get '..tree/binary_space_tree/typedef.hpp:29:32: error: use of class template 'bound::HRectBound' requires template arguments' whith a reference to bound::HRectBound.
< zoq>
I can give you access to the freebsd machine I'm using if that helps
< naywhayare>
hmm, what compiler version? I'll see if I can reproduce it with any of the machines I have first
< zoq>
clang 3.4.1
< zoq>
lets see if it also fails with clang 3.6
< zoq>
yes the same problem
< naywhayare>
no error with clang 3.5; I'm now trying to get a copy of clang 3.6
< zoq>
I'm using clang 3.6 on OSX whitout any errors
< naywhayare>
hmm
< naywhayare>
well, if you give me access to your freebsd machine I can poke around and see what I can find
< naywhayare>
or I can keep trying to find a system with clang 3.4.1 (I'm not having too much luck)
< zoq>
can you send me your public key?
< naywhayare>
hang on, I think I found a system with clang 3.4
< naywhayare>
need to install it first...
< naywhayare>
I can't reproduce it on Fedora 20 with clang 3.4.2
< naywhayare>
okay, I emailed you my public key
< zoq>
okay, thanks I emailed you the credentials
< zoq>
Hopefully it works, the key setting are very restrictive.
< zoq>
If not I can modify the ssh config
< naywhayare>
yeah, works fine; thanks
< naywhayare>
wow that's a lot of warnings...
< zoq>
yeah this warnings about use a struct instead of a class
< naywhayare>
yeah, I thought I fixed that, maybe I didn't push the change
< naywhayare>
I've reproduced the error, though; I'll take a look into it shortly
< zoq>
sounds promising
< naywhayare>
seems like typedef.hpp is never being included; but I haven't figured out why yet
< zoq>
right, I could fix the error by including the 'typedef.hpp' which results in another error
< naywhayare>
hmmm, this is a weird issue
< naywhayare>
it looks like it is trying to include things from /usr/local/include/
< naywhayare>
instead of the correct include path
< naywhayare>
which is the local build directory
< naywhayare>
if I #include <mlpack/core/tree/hrectbound.hpp> then #include <mlpack/core/tree/binary_space_tree/typedef.hpp> at the top of emst_main.cpp, it gives an error referencing /usr/local/include/mlpack/core/tree/hrectbound.hpp
< naywhayare>
I think this will end up being a CMake problem... we need to make sure CMake tells the compiler that the most important include directory is build/include/
< naywhayare>
I'll look into it tomorrow more...
< zoq>
okay, thanks!
< zoq>
you are absolutely right, I removed the old headers from /usr/local/include and now it works ... my bad, sorry for all the unnecessary messages
KTL has joined #mlpack
KTL has quit [Quit: Leaving]
travis-ci has joined #mlpack
< travis-ci>
mlpack/mlpack#183 (master - 80adf7a : Marcus Edel): The build passed.
< cntrtrst>
having a problem with linear_regression. If I feed it data of up to 160 lines it's ok, but beyond that it fails with the message "[FATAL] The model was trained on 4294967295-dimensional data, but the test points in 'testfile.txt' are 4-dimensional!"