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/
freya5859[m] has left #mlpack []
< anjishnu[m]> Hey everyone! Here is my weekly update :) https://iamshnoo.github.io/blog/week-5/
ImQ009 has joined #mlpack
aknd has joined #mlpack
aknd has quit [Remote host closed the connection]
aknd has joined #mlpack
aknd has quit [Remote host closed the connection]
< shrit[m]> rcurtin: I am happy to help, I agree, If I were at your place I would wait for all check to be good before merging
< zoq> anjishnu[m]: Wow, super cool updated, we should definitely look into optimizing the backward pass.
< zoq> anjishnu[m]: Did you make those drawings?
< rcurtin> anjishnu[m]: that's way nicer than any confusing scribbles I have ever written on my whiteboard :-D
< himanshu_pathak[> Hey sakshamb189: can you help me in this http://ci.mlpack.org/job/pull-requests%20mlpack%20memory/5628/
ggalan87[m] has joined #mlpack
< rcurtin> himanshu_pathak[: I haven't looked at the code or anything, but one thing you could do to try and reproduce this is run the tests with valgrind and look at the output:
< rcurtin> $ valgrind --leak-check=full --track-origins=yes bin/mlpack_test -t KernelSVMTest
< rcurtin> you'll want to compile with debugging symbols (-DDEBUG=ON in the CMake configuration), and it might give a lot of output :)
< rcurtin> but it should hopefully be helpful in seeing where the memory access violations are
< himanshu_pathak[> Thanks I will try it :)
jeffin143 has joined #mlpack
< jeffin143> Same error, Although just like you told last time, I resolved the error in previous step of BOOST_INCLUDE_DIRS Not used
< jeffin143[m]> Umm I tried out something let's see if that works
< sreenik[m]> anjishnu: That was a great effort! I mean, I have never really drawn a computation graph to debug anything before. Also, the rate at which you are discovering stuff (that aren't correctly done) is quite alarming :D
pie3 has joined #mlpack
< anjishnu[m]> > anjishnu: Did you make those drawings?
< anjishnu[m]> Yep, all of them. :)
< anjishnu[m]> > anjishnu: that's way nicer than any confusing scribbles I have ever written on my whiteboard :-D
< anjishnu[m]> thank you :P
jeffin143 has quit [Remote host closed the connection]
< anjishnu[m]> > anjishnu: That was a great effort! I mean, I have never really drawn a computation graph to debug anything before. Also, the rate at which you are discovering *stuff* (that aren't correctly done) is quite alarming :D
< anjishnu[m]> Well, as long as we find them... :)
< sreenik[m]> :)
< sreenik[m]> I'd suggest you to open a patch for the batchnorm (would be a one line change I guess)
< anjishnu[m]> Should I just mention the change directly in #2474, because that is already a pretty major refactor of the particular layer?
< sreenik[m]> Oh yes, that would be better
< rcurtin> eddelbuettel[m]: ping, does this communication channel work? :)
< zoq> jeffin143[m]: Looks like a new issue to me?
pie3 has quit [Read error: Connection reset by peer]
pie3 has joined #mlpack
jeffin143 has joined #mlpack
< jeffin143> zoq[m] : are you there
< jeffin143> build\include\mlboard\filewriter\crc_impl.hpp(72,0): Error C7525: inline variables require at least '/std:c++17'
< RyanBirminghamGi> jeffin143: what branch are you getting that on?
< jeffin143> RyanBirminghamGi : Ignore the error due to rmdir , I removed it
< jeffin143> walragatver[m] : Are you there ?
< jeffin143> > what branch are you getting that on?
< jeffin143> Image-support
< RyanBirminghamGi> Do you intend to have mlboad require c++17?
< jeffin143> walragatver[m] RyanBirminghamGi : Last three days updates -> I was working on hsitogram support , I tried to replicate np.histogram() , using `std::lowerbound` but there was small implmentation struggle using that
< jeffin143> I am trying to mirror it
< jeffin143> > Do you intend to have mlboad require c++17?
< jeffin143> No I want it to be C++11
< jeffin143> Just like our other repos
< RyanBirminghamGi> Looks like that line is static on master, but inline on your branch. Did you update the crc impl from somewhere?
< jeffin143> RyanBirminghamGi : Yes I did update it on my branch
< jeffin143> RyanBirminghamGi : Sorry
< jeffin143> My bad
< jeffin143> I removed it, we don't need inline
< jeffin143> let's stick to static
< walragatver[m]> jeffin143 : So great
< walragatver[m]> I think we can merge image support this week
< jeffin143> Definitely,
< jeffin143> I would want that
< jeffin143> It has been log
< jeffin143> long*
< walragatver[m]> Hey jeffin143: I think we should make different folders for different files like util for util files
< walragatver[m]> similar to summary writer file
< jeffin143[m]> Won't it be too much for a small project ??
pickle-rick[m] has joined #mlpack
< walragatver[m]> I think it's better to keep different folders might help in future.
< jeffin143[m]> Ok , is it ok if I open a new pull request for it ???
< jeffin143[m]> I don't want to break anything currently
< jeffin143[m]> Alos I might have to tinker cmake a little bit may be too
< jeffin143[m]> If that's ok with you ?
< walragatver[m]> Okay fine
< jeffin143[m]> @walragatver:matrix.org: also my company contacted me and they would be going ahead with the offer
< jeffin143[m]> Most probably by 7th of August I would have to onboard
< jeffin143[m]> But I guess I can still devote 40 hours a week due to wfh policy
< walragatver[m]> Jeffin143: Okay good
< jeffin143[m]> These are all tentative dates , Nothin is sure , they just informed me over phone
< jeffin143[m]> And Nothing on paper
< jeffin143[m]> So I think it would take much more time that I have mentioned
< walragatver[m]> Okay
< jeffin143[m]> walragatver: any thing else , on your list ?
< walragatver[m]> jeffin143: What are the things you targeting in the second phase?
< jeffin143> May be by tomorrow morning, I will open the pr for histogram ( I am not sure about the math, I tried debugging but didn't work out, so you have to take a look)
< jeffin143> I am suspecting 4 days for it
< jeffin143> Then probably pr curve another 4 days
< jeffin143> Then embedding support another 7 days
< jeffin143> Once all the three of this done, we will sit and design the ensmallen callback, and at least start with the implementation before second phase review
< jeffin143> After these, I think so mlboard is more or less ready with the major ideas we had in our list, May be ryan or zoq could use or test them and give us any improvements, I am sure they have it
< jeffin143> I am not sure, If I missed out anything
< walragatver[m]> jeffin143: Okay fine
< walragatver[m]> The plan looks good
< walragatver[m]> I would give your reviews as soon as possible on the image support PR
< jeffin143> Thanks
k3nz0_ has quit [Remote host closed the connection]
k3nz0_ has joined #mlpack
< walragatver[m]> jeffin143: Hey can you add the examples you added as examples in image.md as test cases for testing?
< jeffin143> ok
< jeffin143> you mean plot a image summary ?
< jeffin143> write ?
< jeffin143> but we can't verify whether it is correct or not , since crc32 is not reversible ? if I am not wrong
< himanshu_pathak[> Hey rcurtin: Can you please answer this comment https://github.com/mlpack/mlpack/pull/2456/files#r451773379 :)
< walragatver[m]> jeffin143: Yeah you are correct but I want those testcases to ensure that those examples keep working in future while we keep developing project
< jeffin143> walragatver[m] : Ok
< jeffin143> I will do it
< walragatver[m]> Also MLpack's data::Load utility is not working properly on my system so I am not able to verify those examples that's it.
ImQ009 has quit [Quit: Leaving]
< jeffin143> Its because of bug in stb image did you try with other images too ?
< jeffin143> I think os stb doesn't work properly with 4 channel
< jeffin143> may be
< jeffin143> Did you try with other images
< walragatver[m]> jeffin143: can you look in usr/local/include/mlpack/core/data folder and check whether load_image.cpp.o file exists?
< walragatver[m]> In your system
< jeffin143> ok just a second
< jeffin143> there are no .o files
< jeffin143> everything is .h
< jeffin143> load.hpp load_arff.hpp load_arff_impl.hpp load_csv.hpp load_image_impl.hpp load_impl.hpp load_model_impl.hpp load_vec_impl.hpp
< walragatver[m]> jeffin143: Okay I am getting undefined reference to mlpack::data::LoadImage()
< walragatver[m]> I checked out and saw that it's in load_image.cpp in data folder in mlpack's repo
< walragatver[m]> It's getting built
< jeffin143> I guess that will libmlpack.a right ?
< walragatver[m]> I installed that build but looks like it's not detecting it
< jeffin143> may be rcurtin[m] would be able to ans that ,
< jeffin143> May be there could be a bug
< walragatver[m]> Yeah might be so I would recommend to add those examples as testcases
< jeffin143> ok
< walragatver[m]> jeffin143: Looks like it's a linking error
< walragatver[m]> When I added build/lib using -L option it worked
< walragatver[m]> jeffin143: I am now getting mismatch between auxiliary memory and requested size after running
gtank___ has joined #mlpack
< walragatver[m]> It's from Mat::init()
kyrre has joined #mlpack
vansika__ has joined #mlpack
< jeffin143> walragatver[m] : Can you try some other image ?
< jeffin143> stb doesn't work well with 4 channels
jeffin143 has quit [Remote host closed the connection]
pie3 has quit [Quit: Leaving]