rcurtin_irc changed the topic of #mlpack to: mlpack: a scalable machine learning library (https://www.mlpack.org/) -- channel logs: https://libera.irclog.whitequark.org/mlpack -- NOTE: messages sent here might not be seen by bridged users on matrix, gitter, or slack
System123 has joined #mlpack
System123 has quit [Ping timeout: 268 seconds]
<jjb[m]> ryan tested the PR using the example in the issue ticket. Works splendidly. Need to remove two debug lines.
AshutoshKumar[m] has joined #mlpack
<jonpsy[m]> welcome Ashutosh Kumar
<AshutoshKumar[m]> Thanks
<jonpsy[m]> Would you like to introduce yourself?
<AshutoshKumar[m]> ok
<AshutoshKumar[m]> So I am Final Year BTech Student at MNNIT Allahabad
<AshutoshKumar[m]> CSE
<AshutoshKumar[m]> I have interest in Deep Learning, BlockChain
<AshutoshKumar[m]> I begining with open source and would like to start with mlpack because it aligns with my interests
<jonpsy[m]> Have a look at [CONTRIBUTING.md](https://github.com/mlpack/mlpack/blob/master/CONTRIBUTING.md)
<jonpsy[m]> also, today we have a meet. So feel free to join and ask your question
<jonpsy[m]> * also, I guess today we
<AshutoshKumar[m]> Sure
<jonpsy[m]> zoq: Remember the rocket injector notebook ? Tried running today, had a runtime error + the notebook was loading indefinetely and there's a good chance it's leaking memory. Let's talk about it in meet
<zoq[m]1> <jonpsy[m]> "zoq: Remember the rocket..." <- Hm, I tested it like a week ago, worked fine.
<rcurtin[m]> ahh, unfortunately I have an in-person company meetup over lunch today so I won't be able to make the video chat :(
<zoq[m]1> I have a clash with another meeting as well, not sure I can make it today.
<shrit[m]> Same here
<zoq[m]1> <zoq[m]1> "Hm, I tested it like a week ago,..." <- Looks like some problem with the sos kernel, will take a closer look. Why do you think there is a memory issue?
<jonpsy[m]> because when i tested memory hit 2.1 GB / 2\
<jonpsy[m]> s/\/ GB./
<zoq[m]1> jonpsy[m]: I think this is what I see locally as well, that it hits 2GB quite fast.
<jonpsy[m]> yeah, but i dont recall that happening previously
<jonpsy[m]> <zoq[m]1> "Looks like some problem with the..." <- so this error was "waiting for something" in runtime which never finished executing. Smells like memory leak
<zoq[m]1> jonpsy[m]: This could be anything.
<zoq[m]1> this is what I get
<jonpsy[m]> tracemalloc
<jonpsy[m]> yeah
<jonpsy[m]> "coroutine 'ZMQSocketChannel.get_msg' was never awaited" this is what i meant by "waiting".
<zoq[m]1> Sure, but this is more like some strange issue with the sos kernel, nothing on the ens/mlpack side.
<jonpsy[m]> that's true, because we hadn't faced this before
<zoq[m]1> At least we don't allocate any memory here, we just include a header file.
<jonpsy[m]> that, and i also wanted to talk about the "Python-C" utils in our examples repo. Should we ditch them for SoS, the extra code we write to manipulate our data and send to Python is super cumbersome
<jonpsy[m]> that block of code really feels out of place
<zoq[m]1> If you like you can update the notebooks, but having the python bindings allows us to run the notebooks outside a notebook env.
<jonpsy[m]> True, but I guess we should consider the effort vs reward ratio
<jonpsy[m]> Python's C-API is very error prone, and formatting C++ data to be compatible for Python C-API makes it kinda hard to grasp, given it isn't even related to the actual code.
<zoq[m]1> True, in some cases, the Python-C API is not as clean, as I said we can have both worlds, if you like to move all the python code into the sos kernel, go for it.
<zoq[m]1> But at least at this point I don't see a reason to remove the existing utils, maybe we can remove them later if we found a better way.
<jonpsy[m]> I guess, we can make current utils "depreceated"
<jonpsy[m]> and encourage future dev's to use SoS kernel, as for the existing which use utils, we could provide an additional code block doing it in SoS kernel.
<zoq[m]1> and just say you have to run everything in a notebook?
<jonpsy[m]> Ah, I see. So only for the notebooks then
<zoq[m]1> I mean I'm fine to say you have to use a notebook, but it would be nice to run the examples outside a notebook as well.
<jonpsy[m]> I see your point now. So, you're saying users should get plots even in the cpp examples. Correct?
<zoq[m]1> I guess, we could also say, you have to run the python script separately.
<jonpsy[m]> > I see your point now. So, you're saying users should get plots even in the cpp examples. Correct?
<jonpsy[m]> have i gotten it correct?
<zoq[m]1> yes
<jonpsy[m]> Okay, so for the notebook we should prefer SoS kernel for plots. But for existing cpp examples, we can go with whats currently there
<zoq[m]1> Hello everyone, mlpack video meet in about 30 minutes, see https://www.mlpack.org/community.html for more details. Some of us can't make it today, but you can use the zoom room anyway.
aakashi2001 has joined #mlpack
aakashi2001 has quit [Read error: Connection reset by peer]
System123 has joined #mlpack
System12_ has joined #mlpack
System123 has quit [Ping timeout: 252 seconds]
System12_ has quit [Ping timeout: 268 seconds]
System123 has joined #mlpack
System12_ has joined #mlpack
System123 has quit [Ping timeout: 260 seconds]
System12_ has quit [Remote host closed the connection]
System123 has joined #mlpack
System123 has quit [Ping timeout: 265 seconds]