07:08 < TanviAgarwalGitt> @shikhar goswami try to go through this link

07:08 < TanviAgarwalGitt> https://www.mlpack.org/doc/mlpack-git/doxygen/build_windows.html

07:36 < TanviAgarwalGitt> @rcurtin sir can you help me in installation of mlpack on windows m not able to get through it?

08:16 < jenkins-mlpack2> Project docker mlpack nightly build build #628: NOW UNSTABLE in 3 hr 2 min: http://ci.mlpack.org/job/docker%20mlpack%20nightly%20build/628/

09:48 < SaraanshTandonGi> I am very very interested in expanding mlpack library with more and more modules for gsoc. I have put up a couple of issues on github. Can someone take a look at them and guide me as to how to proceed?

09:49 < SaraanshTandonGi> I have experience in pytorch as I have been using it since last year for academic research.

09:49 < SaraanshTandonGi> I tried to get into existing issues like the activation functions one but most of them are already taken.

09:52 < SaraanshTandonGi> So I invested this time in trying to understand the flow of ann. With the help of @zoq and @prince776

09:53 < SaraanshTandonGi> and @gaurav-singh1998 I was able to understand it. So now I want to start contributing.

09:56 < anjishnu> Hi everyone! I am Anjishnu. I am interested in the GSoc project idea "Improvisation and Implementation of ANN Modules"

09:56 < anjishnu> How should I get started with the idea?

10:08 < Shikhar[m]> <TanviAgarwalGitt "@shikhar goswami try to go throu"> Hey, I looked at this link. I think The way I installed mlpack(using vcpkg) installs all the required libraries. I think the build environment sections helps them install individually. Correct me if I’m wrong. But once I have installed it. I don’t see any link and can’t find now what I have to do.

12:50 < SriramSKGitter[m> @shikhar : Did you have a look at https://www.mlpack.org/community.html ? It should have something of use :)

12:52 < SriramSKGitter[m> @ajnishu : You could start by getting familiar with the ann module of mlpack, or perhaps looking at https://github.com/mlpack/models for real-world usage.

14:13 < temporary_sreeni> Hey anjishnu, you can get yourself familiarised with the ANN module by building a few models and running them. Try cloning the *models* repo.

14:16 < temporary_sreeni> You can then solve a few interesting issues or try and hunt down the features you’d like to implement in mlpack

15:55 < TanviAgarwalGitt> @shikhar goswami i am stuck up in the middle process of this link, I can't help here if you could help me with this It woudl be great.

16:05 < AbhiSaphire> Hey folks, I am having problem while building mlpack library components on my ubuntu system. While using make command to build library components I am getting this error -->>"make: *** No targets specified and no makefile found. Stop". How can I resolve this issue ??#mlpack

16:12 < metahost> AbhiSaphire: you should find this useful https://www.mlpack.org/doc/mlpack-git/doxygen/build.html

16:19 < SaraanshTandonGi> @prince776 did you get a chance to look at the code?

16:20 < SaraanshTandonGi> While training with two convolutional layers on mnist, I'm getting a very low accuracy. It starts off with 8 percent in the first epoch

16:21 < SaraanshTandonGi> Does anyone have any idea why that might be happening?

16:24 < shrit[m]> @Saraansh Try glorot initialization instead of random

16:28 < GauravSinghGitte> Hey, @saraansh1999 I am not sure what might be happening but substitute the value of `MaxIterations` of the optimizer as the same number of data points in the datasets and see what happens because it might be happening that the optimizer might not be reading the entire dataset, you have a look at the issue #2226.

16:28 < GauravSinghGitte> (edited) ... you have ... => ... you can have ...

16:29 < SaraanshTandonGi> It will give an error, as every value above 23 for some reason is not working.

16:32 < AbhiSaphire> SaraanshTandonGi You can have a look on these solutions, I think it might help https://datascience.stackexchange.com/questions/38604/too-low-accuracy-on-mnist-dataset-using-a-neural-network

16:51 < SaraanshTandonGi> `FFN<NegativeLogLikelihood<>, RandomInitialization> model`

16:51 < SaraanshTandonGi> Which of the constructors of FFN will be called if I write this command.

17:13 < TanviAgarwalGitt> can't we install cmake via visual studio nuget managaer?

17:53 < himanshu_pathak[> Hey zoq sakshsham I was looking at the GSOC project of Essential deep learning module and It includes idea of implementing WGANGP which is already implemented.

17:54 < himanshu_pathak[> * Hey zoq saksham I was looking at the GSOC project of Essential deep learning module and It includes idea of implementing WGANGP which is already implemented.

17:55 < himanshu_pathak[> Only the implementation of PACGN is left

17:55 < himanshu_pathak[> * Only the implementation of PACGAN is left

17:58 < SaraanshTandonGi> I want to debug the internal files of mlpack. So everytime I add a print statement in for eg a visitor, do i have to make the whole of mlpack again?

18:01 < himanshu_pathak[> Instead of using print use gdb It will be more effecient and you have to build ann module only in case of visitor

18:02 < GarvTambiGitter[> Hi @matrixbot i want to interact with you..

18:02 < GarvTambiGitter[> I want to be a part of GSoc20 and want to contribute in your project because i want to select in GSoC

18:03 < SaraanshTandonGi> @himanshupathak21061998 can you help me out please

18:03 < SaraanshTandonGi> I've been stuck for a while now

18:04 < himanshu_pathak[> Can yu explain you problem

18:04 < SaraanshTandonGi> I'm getting an as_scalar() error in Train function somewhere when I put the value of ITERATIONS_PER_CYCLE (maxIterations for optimizer) > 23

18:04 < SaraanshTandonGi> the dataset is mnist

18:05 < SaraanshTandonGi> > https://drive.google.com/drive/folders/1bHwfqIQcidgX8sOiwoMFdQWkWA9G4DnB?usp=sharing

18:05 < SaraanshTandonGi> .

18:05 < SaraanshTandonGi> this link has the file to be run and the data

18:05 < SaraanshTandonGi> i used gdb and saw that the weightsetvisitor are being reached for all layers

18:06 < himanshu_pathak[> Ok I will tr to look into this let me see

18:06 < SaraanshTandonGi> I'm trying to figure out where its going wrong. Thanks.

18:07 < SaraanshTandonGi> This is the error

18:08 < SaraanshTandonGi> It'll be quite embarassing if its some stupid mistake :P

18:12 < GarvTambiGitter[> Please anyone guide me...it would be great help..

18:13 < himanshu_pathak[> Saraansh It needs some diging I can't able to get this why it is happening. Sorry for not able to help you

18:13 < himanshu_pathak[> Hey GarVtambi Try to look at https://www.mlpack.org/gsoc.html

18:15 < himanshu_pathak[> And look try to sove some good first issue in mlpack repo

18:15 < himanshu_pathak[> * Try to solve some good first issue in mlpack repo

18:16 < SaraanshTandonGi> @zoq @rcurtin are you guys free to look into this?

18:17 < SaraanshTandonGi> > `himanshu_pathak` Saraansh It needs some diging I can't able to get this why it is happening. Sorry for not able to help you

18:17 < SaraanshTandonGi> It is reproducing on your system too right>

18:17 < SaraanshTandonGi> (edited) ... too right> => ... too right?

18:21 < TanviAgarwalGitt> https://www.mlpack.org/doc/mlpack-git/doxygen/build_windows.html after following this link I found I am not able to proceed with armadillo dependency can anybody resolve this I am trying it since yesterday but no output and i am lagging behind.

18:25 < himanshu_pathak[> Actually Saraansh I was trying to rebuild mlpack on my pc that's why I can't able to run it when the build will complete I will try it. I thought I will find something while reading but was not able find anything.

18:27 < himanshu_pathak[> Tanvi can you tell us What error are you facing.

18:42 < SaraanshTandonGi> I think i found the line where the error is occuring using gdb

18:42 < SaraanshTandonGi> https://github.com/mlpack/mlpack/blob/225ec4e951a5333aa6c84db1490c76e9ca873977/src/mlpack/methods/ann/layer/max_pooling.hpp#L36

18:42 < SaraanshTandonGi> How can i resolve this?

18:42 < SaraanshTandonGi> and why is this error affected by maxIterations?

18:47 < SaraanshTandonGi> I saw this error in some example code that I was trying to run too. It got resolved once I replace arma::find and arma::max functions with arma::index_max

18:48 < SaraanshTandonGi> But find has been given 1 as its second input. Why should this be happening

18:56 < himanshu_pathak[> 1 is their to return at most 1 indices of input.max() == input

18:58 < himanshu_pathak[> But I can't able to get it whta

18:59 < himanshu_pathak[> Why it is happening because of max iteration

18:59 < SaraanshTandonGi> > `himanshu_pathak` 1 is their to return at most 1 indices of input.max() == input

18:59 < SaraanshTandonGi> exactly, so this error should not come up.

19:01 < himanshu_pathak[> Yes, it should not be the case here. May be you should open a issue for this

19:02 < SaraanshTandonGi> I'm just scared that it might turn out to be something stupid. :P

19:05 < himanshu_pathak[> Don't be scared we are human we can do mistakes😃

19:06 < SaraanshTandonGi> @himanshupathak21061998 can you take a look at the two issues I have opened already?

19:06 < SaraanshTandonGi> about functionalities.

19:24 < himanshu_pathak[> Yes they looks good I am not sure instance norm but we don't have implementation of spatial transform

19:24 < himanshu_pathak[> * Yes they looks good I am not sure for instance norm but we don't have implementation of spatial transform

20:04 < anjishnu> SriramSKGitter[m and sreenik[m] yep, I am looking at the ANN module and the basic models today. I am also trying to hunt down some features and will probably have something more concrete to say about this tomorrow.

20:22 < rcurtin> ha... I celebrated yesterday when mlpack was added to the Julia package repository because I thought that I was done

20:22 < rcurtin> and this morning I found out that mlpack algorithms that use BLAS/LAPACK segfault on 64-bit systems

20:22 < rcurtin> sigh ...

20:23 < zoq> every algorithm?

20:24 < rcurtin> knn doesn't use BLAS/LAPACK so it's fine for now :)

20:24 < rcurtin> I believe I figured out the issue, but it was tricky

20:25 < zoq> Interesting, I think the conda package also uses knn for testing if the package worked, but we should change that as well.

20:25 < rcurtin> the Armadillo version being distributed via the Julia package system links against OpenBLAS, but on 64-bit systems they have renamed all the symbols so they terminate in `64_`, so, e.g., `dgeev64_` not `dgeev`

20:25 < zoq> ...

20:25 < rcurtin> we've got that taken care of just fine already

20:25 < rcurtin> but the part that didn't get noticed is that their OpenBLAS package is using 64-bit integers, while Armadillo is expecting 32-bit integers

20:25 < rcurtin> so ARMA_BLAS_LONG has to be defined... I'm testing the fix now

20:26 < rcurtin> I'm still not clear on why the OpenBLAS symbols are all changed though

20:26 < rcurtin> I don't see any reason to have done that...

20:27 < zoq> so, just a few more hours to get the package working I guess

20:27 < rcurtin> sigh... I hope...

20:27 < rcurtin> plus review time, rebuild time, etc.

20:28 < rcurtin> once I have the fix, I have to push that to the Armadillo Julia package, and once that's reviewed and merged, the mlpack package has to be rebuilt...

20:28 < rcurtin> the folks who maintain the Julia binary packages have been very quick and responsive though, so hopefully not too long

20:28 < rcurtin> much different than the folks who decide on what names are acceptable for Julia packages :)

20:28 < zoq> ahh, that is good :)

20:29 < himanshu_pathak[> Fingers crossed may be this time it will work

20:29 < rcurtin> himanshu_pathak[: yeah, maybe if I make it work I can get back to all the other packaging nightmares that I have :(

20:30 < rcurtin> I have to say, Julia's setup is *way* better than Python's for something like this

20:30 < rcurtin> setuptools is totally opaque and hard to work with for how it compiles Cython bindings

20:30 < rcurtin> but the Julia packages are built directly with g++

21:13 < SaraanshTandonGi> Can someone pls help me out. I have been stuck at this thing and am not able to move forward. #2252

21:13 < SaraanshTandonGi> Sorry for bugging you guys. But i really wanna move forward.

22:20 < rcurtin> SaraanshTandonGi: I know that you want to contribute, but you only opened #2252 *three hours ago*

22:20 < rcurtin> you can see that there are 84 issues open and 76 pull requests open... and that's only on the mlpack repository

22:20 < rcurtin> I'm sure that someone will try and help you when they have time, but that may be quite a while until anyone is available

22:21 < rcurtin> my suggestion would be, if you are stuck on it, try to dig into the code and understand what's going on, or find some other workaround for the issue

22:21 < rcurtin> I'm sorry that I can't give a much better answer than that---in an ideal world, there would always be people to help all the time, but the reality is that all of us work on mlpack in our free time, and that means there's very limited time available