05:53 < AyushSingh[m]> Regarding GSOC ideas 2021, would there be some project in NLP as well?

09:26 < ShahAnwaarKhalid> How do I convert a matrix of type Mat<double> to Mat<unsigned char> for data::SaveImage() function?

09:26 < ShahAnwaarKhalid> I tried Armadillo's .conv_to() function but it gives an error saying " given object doesn't have exactly one element

09:26 < ShahAnwaarKhalid> terminate called after throwing an instance of 'std::logic_error"

13:52 < AvikantSrivasta4> What all previous work done for running MLpack on resources constraint devices or RPi?

13:55 < say4n> Avikant Srivastava you can check out Omar's (shrit on GitHub) work on it. :)

13:56 < AvikantSrivasta4> Alright

16:00 < NippunSharmaGitt> Hi, will there be a project related to the revamping of bindings (#2421) for GSoC 2021 ?

16:03 < rcurtin[m]> Nippun Sharma (Gitter): there could be a binding-related project; not sure yet, it depends on a lot of things (including mentor interest, etc.). personally, I think #2421 could be a good project

16:04 < rcurtin[m]> there is also something else that hasn't been written anywhere yet... the bindings are not threadsafe. specifically, the `IO` object is a global singleton, and as a result it's not possible in some languages to run multiple bindings simultaneously. we found this at my company when attempting to use the Julia bindings in parallel

16:05 < rcurtin[m]> so, that is a thing that needs to be fixed at some point and might be good for a GSoC project (or part of one?)

16:05 < rcurtin[m]> anyway, it is too early to say... we haven't even been accepted as a mentoring organization yet :)

16:05 < kaushal07[m]> Hey ,the Example Zoo idea is quite interesting

16:06 < kaushal07[m]> I would really like to move forward onto it.

16:30 < say4n> Mohomed Shalik : This should be a good place to begin at: https://www.mlpack.org/community.html :)

16:37 < NippunSharmaGitt> @rcurtin thanks for the information.

16:38 < NippunSharmaGitt> I did not know about the bindings not being threadsafe, I will try to reproduce this issue on my local machine regarding this tomorrow.

16:39 < rcurtin[m]> yeah, I am not sure if the issue is reproducible in every language, but it's at least reproducible in Julia :)

16:39 < rcurtin[m]> the results can be very very strange 😃

16:40 < rcurtin[m]> basically, each individual thread will try to modify values of parameters in the `IO` singleton, so whenever `IO` is accessed from inside `mlpackMain()`... `IO` could hold anything, not the intended values

16:40 < rcurtin[m]> there are many different ways that this issue could be solved---I haven't put much thought into what the right solution is (or even what a feasible solution might be)

16:46 < NippunSharmaGitt> I think it will be a good idea to check whether this issue pops up for the rest of the bindings, I will check that. What are some of the ways that this can be solved ? just asking as I still have to get more familiar with this

16:48 < rcurtin[m]> well I am not quite sure... I haven't had a chance to think about it :)

16:48 < rcurtin[m]> in my company we just made a mutex to ensure that only one thread calls an mlpack binding at a time

16:49 < rcurtin[m]> but that is really, really not a great solution; it's just a temporary workaround

16:49 < rcurtin[m]> maybe it's possible to have a super simple solution that would take basically no time at all---I'm not sure

16:59 < NippunSharmaGitt> thanks, I will try to spend some time on this and maybe I can come up with some solution

17:00 < rcurtin[m]> awesome, feel free to open an issue for discussion (or we can keep chatting here, totally your call)

17:09 < NippunSharmaGitt> I think an issue will be better to keep all the discussion recorded in a single place. If it is fine with you, I will open an issue when I know a little more about this so that I can write a good and detailed description, till then if I have any small doubts I can ask here ?

17:16 < rcurtin[m]> yeah, that sounds great to me---if you manage to find a quick way to reproduce it, paste that too. the way we did it internally at my company is way too complicated, but the problem is definitely that `IO` isn't threadsafe