07:14 < RishabhGarg108Gi> Hey @zoq , I tried to serialize a test class (code file -> https://drive.google.com/file/d/1rFT8V4x-QHHQP6knNm09sR-gNGRgwwGC/view?usp=sharing), but it is throwing an error that "static assertion failed: cereal could not find any output serialization functions for the provided type and archive combination." Can you please help me to identify the problem here? Thanks! :D

07:25 < RishabhGarg108Gi> Sorry to bother. The error got resolved. Actually I didn't write the `serialize` function.

11:46 < arcinston[m]> hi guys im new to open source and wanted to know know how do i start

11:47 < arcinston[m]> https://www.mlpack.org/gsoc.html i did go through this and i have coded in c++ which is one of the reasons i selected this project

11:47 < arcinston[m]> any help guiding me would be appreciated

11:50 < RishabhGarg108Gi> Hey @arcinston , you can start with some `good first issue`. It will help you to understand the codebase.

11:50 < RishabhGarg108Gi> If you face any kind of difficulty, feel free to ask here :)

11:50 < arcinston[m]> ill do that thanks man

11:53 < arcinston[m]> do you recommend i do all work on linux or will windows work

11:54 < arcinston[m]> im sorry if this sounds dumb :)

11:56 < zoq> arcinston[m]: You can stick with Windows, the build process is not as straightforward as doing it in Linux but it's possible.

11:57 < zoq> arcinston[m]: I guess the new build tutorial - https://github.com/mlpack/mlpack/pull/2740 should be useful.

12:05 < arcinston[m]> well now im using linux so i dont cause more trouble

12:20 < zoq> arcinston[m]: Great, easier for us to provide support :)

12:20 < RishabhGarg108Gi> Hey @zoq , I tried serialization a few simple classes and also some models.

12:21 < zoq> RishabhGarg108Gi: Sounds good.

12:22 < RishabhGarg108Gi> It turns out that there are two issues here. First is that since in serialization, we are actualling storing the state of the classes, and we majorly focus on the data of the class and we do not require that much information in our config file.

12:23 < zoq> RishabhGarg108Gi: Do you have an example?

12:24 < RishabhGarg108Gi> The second issue is that cereal can only parse the json files that are generated from cereal::JSONOutputArchive. Now the issue with the output of such serialization is very complicated that it is very difficult to understand it and also write something in that format.

12:25 < RishabhGarg108Gi> I can make a small model and its corresponding serialized output for you.

12:25 < zoq> Maybe cereal is not the best approach.

12:26 < zoq> That would be great, just comment on the issue, that way other can take a look as well.

12:26 < RishabhGarg108Gi> Okay, I will post a comment on the issue itself.

12:27 < zoq> Super, thanks.

13:15 < jeffin143[m]> Hi @zoq , do you work on mac ?

13:15 < jeffin143[m]> I can reproduce mac pytest isssue in local too

13:16 < zoq> jeffin143[m]: Hey, sometimes.

13:16 < zoq> jeffin143[m]: Ohh you can?

13:16 < jeffin143[m]> Not sure why you weren't able to

13:16 < jeffin143[m]> zoq (@freenode_zoq:matrix.org): yes the same exact error

13:16 < zoq> jeffin143[m]: What version are you using?

13:17 < zoq> jeffin143[m]: Python, macOS and armadillo would be interesting.

13:17 < jeffin143[m]> Mac os cataina 10.15.5, python 2.7.17

13:17 < jeffin143[m]> How do I check armadillo version ??

13:19 < jeffin143[m]> 9.900.1

13:19 < jeffin143[m]> For arma

13:19 < jeffin143[m]> Also ran it using ctest .

13:20 < zoq> Thanks, I'll try to use the same packages as you.

13:20 < zoq> Yes, I testes ctest as well.

13:20 < jeffin143[m]> Even the test were broken with python 3.5 too

13:20 < jeffin143[m]> I created a venv and then tried too

13:21 < zoq> jeffin143[m]: I'm on Catalina 10.15.7, hopefully that's not the issue.

13:23 < jeffin143[m]> I am really unsure if it depends on the specific version

13:23 < jeffin143[m]> I thought it broke for all Mac users , but strangely on your system it passed , Haha

13:23 < zoq> Strange yes.

13:24 < zoq> If I can reproduce the issue, it would be easier to fix :)

13:24 < jeffin143[m]> Yes , just saw your commits and came here , because I can understand the pain of waiting for the ci to build

13:25 < rcurtin[m]> I borrowed a macbook to reproduce the Julia issue, but I haven't managed to figure out what is going on yet...

13:25 < zoq> Thanks, I thought hey I can't reproduce the issue, it might be an Azure specific macOS problem.

13:26 < rcurtin[m]> At least for Julia the issue appears to be memory-related: it seems that Julia is trying to free some memory that was never allocated... not sure if the Python issue is similar at all

13:27 < zoq> A friend of mine is getting the new one with the M1 chip, so I guess I could borrow his old one as well :)

13:27 < jeffin143[m]> rcurtin (@ryan:ratml.org): how come it suddenly broke 1 day , everything was going well

13:27 < zoq> Probably the same issue.

13:27 < jeffin143[m]> zoq (@freenode_zoq:matrix.org): just came to know that m1 is Beast , but i also saw we can't run docker in M1 currently

13:28 < rcurtin[m]> jeffin143: exactly, I have no idea yet. It seems like Armadillo still handles memory the same, and it doesn't seem to be a Julia issue either

13:29 < jeffin143[m]> rcurtin (@ryan:ratml.org): ahaa I see , let's hope you get to the root easily

13:30 < zoq> jeffin143[m]: more than 16GB of memory would be great, but that's currently not supported.

13:30 < zoq> But maybe mlpack will improve on the memory usage, and 16GB is enough :)

13:30 < rcurtin[m]> jeffin143: yeah, hopefully I can figure it out soon. next step is valgrind...

17:01 < _slack_mlpack_13> Hey everyone!

17:01 < _slack_mlpack_13> I am Sashank from GITAM, Visakhapatnam, India!I am an open-source and ML/DL enthusiast!

17:05 < zoq> Hello Sashank.

17:06 < _slack_mlpack_13> (edited) ... get started? => ... get started here?

17:20 < zoq> So a good starting point is the community page - https://www.mlpack.org/community.html especially the Getting Involved section.

17:20 < zoq> Also in case you are interested in GSoC, we have a page for that as well that has some useful information as well - https://www.mlpack.org/gsoc.html

17:21 < RishabhGarg108Gi> Whoa @zoq , your approach that you mentioned [here](https://github.com/mlpack/mlpack/issues/1254#issuecomment-736542011) was really nice. I didn't think of it that way. I made a really basic prototype of what you said and it looks good enough to me.

17:22 < zoq> RishabhGarg108Gi: Excited to see this at some point working.

17:23 < RishabhGarg108Gi> It is still no where near to the actual thing, but I wanted to know if I am heading in the right direction. So, I will show you the prototype in a couple of minutes here. I don't think spamming there would be a good idea until we get to some actual thing. :)

17:24 < zoq> Works for me, I have to step out for some minutes, but once I get back I'll take a look.

17:25 < RishabhGarg108Gi> Great. Its fine, you can finish your other work, I am not in any hurry :D

17:25 < _slack_mlpack_13> Thanks zoq

17:32 < RishabhGarg108Gi> @zoq , [Here](https://drive.google.com/file/d/10E8t0RxzPEopqHt1sqHznlq-qMMTDeZX/view?usp=sharing) is the code that I wrote. And [this](https://drive.google.com/file/d/1VYIyNYyiMBhqAlFnGUGcOMfM7KQ6qUG0/view?usp=sharing) is the config file it generated. Although my code is a bit messy, but I tried to make it as much readable as possible. Wishing you luck to read it :p. Thanks.

21:11 < zoq> RishabhGarg108Gi: I think this looks really readable, we might have to find an approach to store values other than pairs of string int.

21:12 < zoq> RishabhGarg108Gi: So we can stick multiple name parameter pairs into vector<layer<int>> network;