ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
< jonpsy[m]> Ryan Birmingham (Gitter):
NitishChaturvedi has left #mlpack []
bash[m] has joined #mlpack
Agzam has joined #mlpack
Agzam has quit [Quit: Connection closed]
ImQ009 has joined #mlpack
onoplihand1975[4 has joined #mlpack
< onoplihand1975[4> For a long time I wanted to try something new for you URL http://v.ht/S7keg
onoplihand1975[4 has left #mlpack []
< jonpsy[m]> is it cool if i disable appveyor parallel build?
< jonpsy[m]> appveyor is exiting with ```command exited code 1 ```, so i took the entire log and matched it with MSVC Code for errors(C1XX, C2XX) none of them matched.
< jonpsy[m]> * is it cool if i disable appveyor parallel build? (only in PR for checking)
< zoq> jonpsy[m]: Don't think you can change the setting. Also not sure I understand the problem.
izlacociredef has quit [Quit: Connection closed]
< RishabhGargGitte> Hey all, I had sent a mail regarding a GSoC project 1 week ago and I haven't got a response to that. So, I just wanted to confirm if it was received or it ended up in spam or something. Or if it is because none had a chance to reply to it, then its okay. But, if its the first case and it has not been received then shall I resend it ?
< rcurtin[m]> you can check the list archives to see if it went through:
< rcurtin[m]> if you see your message in the archives, then it successfully sent to the list
< rcurtin[m]> whether or not it ended up in people's spam folders is a function of what mail service they are using (i.e. it has nothing to do with the list itself)
< zoq> It was the tree ensemble mail right?
< RishabhGargGitte> Yes @zoq . I can find it in the archives too.
< jonpsy[m]> > jonpsy: Don't think you can change the setting. Also not sure I understand the problem.
< jonpsy[m]> its reg. the NSGA2 error, appveyor exits with no error messages its blowing my mind :)
< zoq> Yeah I got that one, I can give you a general response later, unless our tree experts get to it first :)
< RishabhGargGitte> Thanks @zoq :-). Btw, I am guessing by "tree expert" you mean Ryan ?
< rcurtin[m]> presumably... :) I'll get to it when I have a chance
< zoq> jonpsy[m]: Sure, but I don't see what the parallel setting has to do with the error.
< rcurtin[m]> I have 27 flagged emails to handle right now, so we'll see when `$RAND % 27` returns the index of your message :)
< zoq> rcurtin[m]: RishabhGargGitte: :)
< RishabhGargGitte> Thanks @rcurtin , no hurry :)
< jonpsy[m]> > jonpsy: Sure, but I don't see what the parallel setting has to do with the error.
< jonpsy[m]> i thought maybe because its running in multicore, the error is getting hidden somewhere. SO, if I run single core my error will be just below the build failure
< RishabhGargGitte> Take your time.
< rcurtin[m]> 👍️
< rcurtin[m]> jonpsy: sometimes debugging AppVeyor can be... not a lot of fun 😭
< rcurtin[m]> the builds do have a 90 minute build limit, so if it hits that, that could be a reason that it terminates with seemingly no error
< jonpsy[m]> already have a headache bcz of it :(
< jonpsy[m]> > the builds do have a 90 minute build limit, so if it hits that, that could be a reason that it terminates with seemingly no error
< jonpsy[m]> fails in 41 mins xD
< rcurtin[m]> ahh :(
< zoq> jonpsy[m]: Ahh you mean the build in parallel setting, I thought you are talking about that we build ensmallen in parallelwith different compilers.
< jonpsy[m]> yh, needless to say that didnt work either
< zoq> Did you test to connect via ssh?
< zoq> That is proably the fastest way to debug the issue.
< jonpsy[m]> i have to do it via windows RDP if I'm right?
< jonpsy[m]> > the builds do have a 90 minute build limit, so if it hits that, that could be a reason that it terminates with seemingly no error
< jonpsy[m]> this applies to my MOEAD PR though, the build takes 1h:30 mins and abruptly exits. Any chance I can increase build time?
< zoq> jonpsy[m]: You can use RDP or ssh.
< jonpsy[m]> correct me if im wrong, RDP is for windows VM and ssh for is for Linux VM?
< zoq> jonpsy[m]: The appveyor team already increased the build time for us :)
< zoq> jonpsy[m]: Ahh you are right
< zoq> jonpsy[m]: Maybe another solution is to test it locally?
< jonpsy[m]> i dont code in windows, so i left it as my last ditch effort
< zoq> jonpsy[m]: I can start a VM later and see if I can reproduce the issue.
< rcurtin[m]> you can at least rdp from a Linux system
< rcurtin[m]> and then kind of do things from the Windows command line or powershell
< rcurtin[m]> the main problem is typing 'ls' and getting 'command not found' over and over again 😃
< jonpsy[m]> hahahaha
< jonpsy[m]> allright, coding in windows it is!
< zoq> jonpsy[m]: Looking at the log, maybe you can make NSGA2 a template class and remove the inline from the impl and see if that at least builds?
< jonpsy[m]> template with no args?
< zoq> jonpsy[m]: I would just put typename MatType = arma::mat in there.
< zoq> It's just a test.
< jonpsy[m]> alrighty
< zoq> I cancle the builds for you.
< jonpsy[m]> gimme a min
< jonpsy[m]> getting a thousand more errors
< jonpsy[m]> it compiles..
< jonpsy[m]> alright pushed
< GopiManoharTatir> Hey Ryan,
< GopiManoharTatir> This is regarding the the reply you sent to mailing channel.
< GopiManoharTatir> In case we are planning to remove boost::spirit in the future and we are also trying to create a data frame like class that will use DataMapper, I would suggest that we should implement a by hand csv parse which can be designed keeping this new class in mind. Does this make sense?
< rcurtin[m]> If I remember right, it's worth taking a look at the issue I linked to---there are a number of suggestions for what to do there. personally, I would suggest finding a lightweight dependency that we can use, instead of having the extra maintenance burden of a by-hand implementation 👍️
< GopiManoharTatir> Okay, I will do some research and can include this in my GSoC idea. As I can see before building something like a dataframe we need to do a bit of other stuff to make it happen. I at least want to get this project started in the gsoc as a kickstart for this idea.
< rcurtin[m]> Gopi Manohar Tatiraju: yep, sounds good 👍️
< GopiManoharTatir> I think we should try to find a simple csv parse rather than implementing one. I went through the one's you listed but you already listed the disadvantages. Do you have anything else in mind that would help? Are we still planning to use them despite their downside?
< GopiManoharTatir> I will research more and update...
< rcurtin[m]> Oleksandr Nikolskyy: I'm glad you asked! it gives me the first excuse to point someone to this writeup we did on ensmallen's design: https://arxiv.org/pdf/2003.04103.pdf
< rcurtin[m]> I think Sections 3.2.1 and 3.2.2 will be the relevant ones there
< rcurtin[m]> oh, actually, just Section 3.2 :) (although hopefully the whole document might be helpful)
< OleksandrNikolsk> Thanks for the quick response! I check this out
< ABHINAVANAND[m]> I was building mlpack on colab and the build is completed but I am unable to include the layers. The colab link is
< ABHINAVANAND[m]> Where am I going wrong.
< jonpsy[m]> Hmm, looks like cereal isn't in your include path?
< zoq> ABHINAVANAND[m]: jonpsy[m]: is Right you have to add it, the same way you did for ensmallen.
< zoq> or stb
< ABHINAVANAND[m]> Oh yes I got it. Now I included cereal and it's compiling.
< jonpsy[m]> zoq the build failed :). I'll try RDP tommorow, given this is not my personal repo, I won't get blocked from using it no?
< zoq> jonpsy[m]: Blocked no.
< ABHINAVANAND[m]> I forgot to remove ensmallen. It's actually someone else's notebook which I was taking as a reference to build mlpack on colab.
< jonpsy[m]> <rcurtin[m] "Oleksandr Nikolskyy: I'm glad yo"> Is there a template to this or something? This is awesome!
< rcurtin[m]> jonpsy: it's an article we wrote up as an 'add-on' to our ensmallen JMLR submission: https://github.com/rcurtin/ensmallen_tr , https://github.com/rcurtin/ensmallen_jmlr
< jonpsy[m]> Too bad I can't write in LaTex. But the format is amazing
< rcurtin[m]> thanks 👍️
< ABHINAVANAND[m]> When I passed `ones(49, 1)` into the meanpooling layer, it gave correct output but when I passed `ones(7,7)` it returned `[matrix size: 0x7]` not sure why. The layer input width and height was specified as 7*7.
< NippunSharmaGitt> Hey @rcurtin, I had a little doubt regarding the mail you sent in reply to my project idea (revamping the mlpack bindings). Sorry that I forgot to mention it in my reply. In the mail, you mentioned that it would be better to take one binding and refactor it in the way that I am planning to. So, should this be automatically generated? I am asking because this is just used to demonstrate the idea so will a handwritten
< NippunSharmaGitt> example work? Also, can you tell what is the best way to share the code along with my proposal?
< AyushSingh[m]> Hello everyone, I am a sharing a draft GSoC proposal. Looking forward for review, specifically in 'Project Timeline' section. I will make more changes in accordance with GSoC proposal format of Mlpack.
vansika__ has quit [K-Lined]
< ABHINAVANAND[m]> The ceil mode in pooling layers does not match with the pytorch implementation.
< ABHINAVANAND[m]> The implementation of ceil_mode in pooling layers may not be correct, or am I missing something.
< zoq> ABHINAVANAND[m]: Feel free to open an issue on GitHub.
< rcurtin[m]> ABHINAV ANAND: I feel like I saw an issue or PR open about this already too, maybe worth searching to see what is open already?
< ABHINAVANAND[m]> A pr is open for failing of backward function in meanpooling.
< ABHINAVANAND[m]> I will open an issue.
< rcurtin[m]> 👍️
ImQ009 has quit [Quit: Leaving]
hasan1101 has joined #mlpack
hasan1101 has quit [Client Quit]