10:44 < kaushal[m]> <kaushal[m] "while trying to run examples "> I am still getting these even after compiling boost rebuilding mlpack from source

11:10 < kaushal[m]> <kaushal[m] "while trying to run examples "> hey Aakash kaushik (Gitter) I am still getting same errors even after compiling boost and rebuilding mlpack from source

12:03 < AakashkaushikGit> Can you let me know which example are you trying to run, i can try it on my machine

12:10 < kaushal[m]> lstm_stock_prediction

12:11 < kaushal[m]> I tried everything u guys told me

12:11 < kaushal[m]> but it doesn't seem to solve my problem

12:12 < kaushal[m]> well I was thinking

12:12 < kaushal[m]> do i need to link the boost module to bashrc profile

12:12 < kaushal[m]> like to linking to compiler paths and linked paths

12:26 < NippunSharmaGitt> Hi. I was building go bindings from source. I passed `-DBUILD_GO_BINDINGS=ON` to cmake and then ran `make`. It builds without any errors. But when I try to `import mlpack.org/v1/mlpack` it is not able to identify the package (I also tried make install but that did not change anything). Is there anything else that I have to do ?

12:27 < NippunSharmaGitt> sorry read it as `import "mlpack.org/v1/mlpack"

12:27 < NippunSharmaGitt> `import "mlpack.org/v1/mlpack"`

14:37 < jonpsy[m]> I'm not a pro, but it perhaps has to do with the PATH

14:38 < jonpsy[m]> Like python needs PYTHONPATH. So if I'm right, it's probably searching for this module in its corresponding PATH variable

14:38 < rcurtin[m]> yeah, if I remember right, GOPATH has to be set right

14:39 < jonpsy[m]> Yeah, in case of mlpack. It automatically sets the PYTHONPATH for you, in case you're wondering why it didn't occur in python

14:40 < rcurtin[m]> PYTHONPATH doesn't get set automatically by the mlpack build process, except when calling `ctest` to run the python binding tests

14:40 < rcurtin[m]> (I think `ctest` will also properly set `GOPATH` for the Go binding tests)

14:40 < rcurtin[m]> it may be easier to test the Go bindings by following https://www.mlpack.org/doc/mlpack-git/doxygen/go_quickstart.html and then dropping the hand-compiled .go files in place...

14:41 < jonpsy[m]> Umm

14:43 < jonpsy[m]> <rcurtin[m] "PYTHONPATH doesn't get set autom"> Correct me if I'm wrong, but docs say "mlpack Python binding should be automatically in your PYTHONPATH and importing mlpack functionality into ... "

14:43 < jonpsy[m]> In README.md

14:43 < rcurtin[m]> if you install, yes:

14:43 < rcurtin[m]> > If mlpack is installed to the system ...

14:44 < jonpsy[m]> with cmake python binds ON + make install, right?

14:44 < rcurtin[m]> yes, right; but if you have just built mlpack in a build directory, then nothing special happens with PYTHONPATH and you will have to set PYTHONPATH manually if you want to use the mlpack Python bindings from the build directory without installing them

14:44 < jonpsy[m]> If I do cmake go bindings ON + make install

14:45 < jonpsy[m]> Will that work the same way?

14:45 < jonpsy[m]> because she tried make install as well

14:47 < rcurtin[m]> I'm not sure how `make install` works for the Go bindings... generally we did not expect users to do that, since it is much easier to install via `go get` (through the mlpack-go repository)

14:47 < rcurtin[m]> if `make install` should work differently for the Go bindings, definitely feel free to open an issue or a PR to fix it 👍️

14:48 < rcurtin[m]> whenever I've used the Go bindings after hand-building them, I have always just set `GOPATH` to include the build directory of wherever I built them

14:49 < NippunSharmaGitt> isn't the GOPATH automatically set while installing ?

14:49 < jonpsy[m]> Thanks for clarifying. Yes, I think it'll be a nice feature to automatically set corresponding PATH(in our case Go) variable when doing make install. I think Nippun Sharma (Gitter) would be interested in that, it's her domain :)

14:50 < rcurtin[m]> setting the system GOPATH during installation doesn't make sense---instead, we probably need to figure out how to install the bindings to whatever the GOPATH is already set to

14:51 < rcurtin[m]> you can take a look at how we do it for python---if I remember right, we take a guess at what the system PYTHONPATH is before calling `setup.py install`

14:52 < rcurtin[m]> and then also provide a CMake option so that the user can specify where they want the Python bindings to be installed

14:52 < rcurtin[m]> let me see if I can find a link...

14:52 < rcurtin[m]> https://github.com/mlpack/mlpack/blob/master/src/mlpack/bindings/python/PythonInstall.cmake

14:53 < rcurtin[m]> so you can see that we use the `PYTHON_INSTALL_PREFIX` CMake variable or `DESTDIR` environment variable if it's set

14:53 < rcurtin[m]> in that case, the install script has to be a separate CMake script because we need to invoke the setuptools installer, but for Go we can probably use more "normal" `install()` CMake commands in the main `CMakeLists.txt`

14:54 < rcurtin[m]> I am not a Go expert so it's always possible that things can be improved; my suggestions may not be the best way possible of course 😃

14:54 < rcurtin[m]> Yashwant wrote the Go support, so you could ping him if you had more questions... I'm about at the limit of my knowledge here 😃

14:54 < jonpsy[m]> Cool, I think all this would be pretty helpful for Nippun Sharma (Gitter)

14:55 < jonpsy[m]> By the way, guess what? I'm adding ZDT Test suite!

14:55 < rcurtin[m]> 👍️ I saw that (didn't have a chance to look into it at all though)

14:56 < NippunSharmaGitt> @rcurtin @jonpsy I am pretty new to GO also.. but the comments seem very helpful.. I will tell if I get any success. Thanks

15:37 < RiturajDuttaGitt> When I try to run the example program of mlpack I got a ton of error specifically stating undefined reference and lastly collect2: error: ld returned 1 exit status

15:37 < rcurtin[m]> Rituraj Dutta (Gitter): what have you done to try to debug the issue?

15:38 < RiturajDuttaGitt> ` g++ -o test_1 test_1.cpp -fopenmp`

15:38 < RiturajDuttaGitt> I've used this command

15:38 < rcurtin[m]> yes, that's how you produced the issue, but what have you done to try to debug it?

15:42 < RiturajDuttaGitt> I was going according to the docs and it didn't mention anything except that mlpack has itself a debugging log which I also I'm unable to compile

15:44 < RiturajDuttaGitt> Is it a linker issue

15:44 < rcurtin[m]> have you looked into what the `undefined reference` errors might mean?

15:46 < RiturajDuttaGitt> undefined reference generally occurs when none of the object files I'm trying to link contains a definition

15:47 < RiturajDuttaGitt> *for a function

16:54 < RishabhGargGitte> Hey @rcurtin , (here)[https://github.com/mlpack/mlpack/blob/3d0e439d499ef672516d2bf9d51c25b7c0dfdeb9/src/mlpack/methods/decision_tree/information_gain.hpp#L85 ] can't we use just a `double` instead of `double []` ? Or is this to leverage the SIMD instructions ? I am at all familiar with SIMD instructions, but if this array is not meant for it, then I guess we can simply replace that with a double accWeights. Let me

16:54 < RishabhGargGitte> know what do you think about it.

16:55 < RishabhGargGitte> **I am not at all familiar with

16:55 < rcurtin[m]> yes, exactly, we use four `double`s so that the compiler can easily generate instructions to perform each step of the loop all at once

16:56 < rcurtin[m]> SIMD is "single instruction multiple data", so you can think of it a little bit like parallelism

16:56 < rcurtin[m]> so, e.g., a SIMD add will add together 8 elements to produce 4 outputs, in (basically) the same time it takes a non-SIMD add to add together 2 elements to produce 1 output

16:56 < rcurtin[m]> but in order to get the compiler to produce SIMD instructions, you have to be pretty careful about the types you are using and how they are aligned

16:57 < rcurtin[m]> so in this case, we need to keep all four elements in the `double[]`, otherwise the compiler will not generate the much faster SIMD instructions

17:01 < RishabhGargGitte> Okay, just to make sure I am following you, if we change `double accWeights[4]` to `double accWeights` then the compiler will not leverage SIMD for the `count`, `count2`, `count3` and `count4. Right ?

17:02 < rcurtin[m]> yes, exactly 👍️

17:02 < rcurtin[m]> if you have some time and interest, SIMD is really cool and worth learning about if you like low-level optimization

17:03 < rcurtin[m]> http://ftp.cvut.cz/kernel/people/geoff/cell/ps3-linux-docs/CellProgrammingTutorial/BasicsOfSIMDProgramming.html is a nice place to start (one of many), though it is somewhat specific to programming a PS3 :)

17:03 < RishabhGargGitte> Okay. Thanks :). It was a completely new concept for me. I will surely explore it a bit more.

17:03 < RishabhGargGitte> Thanks. I will look at it. :+1:

17:44 < kaushal[m]> hey I am trying to build python bindings but it's not able to build it

17:44 < kaushal[m]> its saying pandas not found

17:45 < kaushal[m]> whereas I have already installed pandas

17:46 < RishabhGargGitte> if you are on a UNIX operating system then try installing `python-pandas`. It should probably work.

17:49 < kaushal[m]> ubuntu

17:54 < kaushal[m]> <RishabhGargGitte "if you are on a UNIX operating s"> it says ,can locate module

18:21 < rcurtin[m]> hey Hemal Mamtora , is the cereal version new enough? you can check with `apt-cache policy libcereal-dev`

18:23 < RiturajDuttaGitt> Okay so I have two version of mlpack in my system 3.2 and 3.4 that's why it was getting troublesome. So now how do I remove the other version

18:24 < RiturajDuttaGitt> @matrixbot

18:24 < rcurtin[m]> Hemal Mamtora: I see, are you building in a clean build directory?

18:24 < zoq_> RiturajDuttaGitt: Just remove the header files and lib.

18:24 < rcurtin[m]> maybe remove `CMakeCache.txt` and try configuring with cmake again? based on what you said, CMake should be able to successfully find cereal

18:26 < jonpsy[m]> Hey rcurtin

18:26 < jonpsy[m]> I've always wondered why ```MatType``` is used for population in ensmallen modules. We're sure they're points right?

18:27 < rcurtin[m]> sure, but what type of point? it could be an `arma::Mat<float>`, `arma::mat`, `arma::sp_mat`, ...

18:29 < RiturajDuttaGitt> Thanks @matrixbot finally able to run mlpack :))

18:29 < rcurtin[m]> what is the CMake output that was printed when you tried to configure? I'm not sure the output you pasted is related

18:30 < RiturajDuttaGitt> Just added the flag -L/usr/local/lib/ as mentioned in issue #2709

18:33 < kaushal[m]> <RiturajDuttaGitt "Thanks @matrixbot finally able t"> were u able to run the mlpack programs in the examples folder , like the lstm_stock_prediction

18:33 < rcurtin[m]> is `cereal/cereal.hpp` in `/usr/include/`? Maybe you could just manually specify `-DCEREAL_INCLUDE_DIR=/usr/include/cereal/`...

18:45 < kaushal[m]> hey rcurtin when trying to download all the datasets for the examples folder of mlpack using ./download_data_set.py it shows no tqdm module found , even after installing tqdm module , please help

18:46 < rcurtin[m]> I'm really sorry @kaushal07:matrix.org , I don't have the time to help debug each issue

18:47 < kaushal[m]> ok but I have been facing the issue in running this examples folder for 4 days

18:47 < kaushal[m]> maybe a little suggestion

18:50 < kaushal[m]> hey Hemal Mamtora I am rebuilding mlpack now so I will paste the error log after some time

18:54 < kaushal[m]> well i have python 3.8 and pip3

19:16 < jonpsy[m]> <rcurtin[m] "sure, but what type of point? i"> Thanks for this!

21:53 < brongulusGitter[> Install the `tqdm` module for python?