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/
pie3 has joined #mlpack
< jonpsy[m]>
Hi, my make command is taking too long to compile, help
< jonpsy[m]>
So much so that when I use make -j5 my laptop gets seizure.
< jonpsy[m]>
I get an error "internal compiler killed"
< jeffin143[m]>
Anyone knows how to convert arma rowvec or colvec to std:: vector<double>
< jeffin143[m]>
Or we can't
< jeffin143[m]>
jonpsy: do you have 5 cores ?
< jeffin143[m]>
You can only build the component you are working on using make component_name
< jonpsy[m]>
Yes but it's an .hpp file
< jonpsy[m]>
So if it compiles it recompiles all classes which use it
< jeffin143[m]>
Yes, you are taking about the split pr right ?
< jeffin143[m]>
You can build mlapck_test
< jonpsy[m]>
Yes
< jeffin143[m]>
That is much faster
< jonpsy[m]>
BTW I have 4 cores
< jonpsy[m]>
And 8 threads
< jeffin143[m]>
Mlpack does take quite a few min to build approx build time is 1 hr ,
< jeffin143[m]>
You cannot dodge that
< jonpsy[m]>
WTF
< jonpsy[m]>
That's huge
< jonpsy[m]>
At first time right?
< jeffin143[m]>
Yes only first time
< jeffin143[m]>
Due to catching it take very less time
< jeffin143[m]>
Subsequently
< jonpsy[m]>
Ahh good, yes
< jonpsy[m]>
Thing is I've used make -j7 at times(on another repo)
< jonpsy[m]>
But none have given it seizures of the level mlpack has been giving it recently
< jeffin143[m]>
-j7 is very huge ,your system will hang
< jeffin143[m]>
Like you need very powerful system
< jeffin143[m]>
J is for the core
< jonpsy[m]>
Dang, I took it for threads
< jeffin143[m]>
Yes , so may be build using Make
< jeffin143[m]>
Deafult is -j4
< jeffin143[m]>
That would suffice
< jonpsy[m]>
Oh that's exactly how much I have
< jeffin143[m]>
It would build once , and the subsequently you can work upon with the catch
< jeffin143[m]>
Cache
< jonpsy[m]>
Yeah I see. But still, it didn't load this much before (when I first started this PR) as it's doing now
< jonpsy[m]>
Maybe I'll try make clean and rebuild.. ¯\_(ツ)_/¯
ImQ009 has joined #mlpack
< jeffin143[m]>
Yes , all the best :)
< jonpsy[m]>
Thanks for the help!
< zoq>
jeffin143[m]: You should be able to use arma::conv_to over the rows something like: for (size_t i = 0; i < V.n_rows: ++i) N[i] = arma::conv_to<std::vector<type>>::from(V.row(i)); where V is of type std::vector<std::vector<type>> and N is std::vector<std::vector<type>> N(V.n_rows);
< zoq>
jonpsy[m]: Building template code takes a lot of memory, so make sure you have enough.
< jonpsy[m]>
Yep, learnt it the hard way
< zoq>
You could increase your SWAP as well, but I guess building with less cores is faster, not sure.
< jonpsy[m]>
Building with less cores is stable
pie3 has quit [Remote host closed the connection]
< kshitijandmojoGi>
can i use visual studio for contributing??
< zoq>
kshitijandmojoGi: Visual Studio should be able to load a CMake project.
< kshitijandmojoGi>
do i need to install some extension for that ??
< rcurtin>
I think that I have my scripts set up right so that when it merges, the release process will happen automatically
< abernauer[m]>
I might give it a look though my shell scripting skills are not my strong suit.
< rcurtin>
the PR doesn't have any of the release scripts in it (those are in the mlpack-bot repo), this is just the updates to the ensmallen repo that are necessary for the release. it just needs to be approved by two repository maintainers and then it can be merged
< abernauer[m]>
Well I looked it over looks fine to me approved though I don't have those permissions.
< zoq>
jonpsy[m]: Easy enough to put a template around it, since it's alraedy using the continues element access.
< jonpsy[m]>
Yeah, that we could do, but it ain't "checkmatrice" Then ;)
< jonpsy[m]>
Maybe, "checkmatrice&vector" Or something along those lines (I'm not good at naming)
< zoq>
jonpsy[m]: a vector is just a 1xn matrix :)
< jonpsy[m]>
Ah
< jonpsy[m]>
Mind blown
< zoq>
jonpsy[m]: Ahh, slice returns a subview, try F(i) instead of F.slice(i)
< rcurtin>
thanks himanshu! hopefully you are having some success tracking down the memory issue?
< rcurtin>
I like the gif :)
< himanshu_pathak[>
> thanks himanshu! hopefully you are having some success tracking down the memory issue?
< himanshu_pathak[>
Yup trying some changes may be that will work thanks for reading :)
< rcurtin>
sometimes these things can be super hard to track down
< rcurtin>
let me know if you run out of ideas and I can try to take a look at the code, but maybe saksham might have better ideas (I'm sure he's more familiar with the code at this point)
< himanshu_pathak[>
> let me know if you run out of ideas and I can try to take a look at the code, but maybe saksham might have better ideas (I'm sure he's more familiar with the code at this point)
< himanshu_pathak[>
Sure I will ask for a review on my pr thanks for helping.
< rcurtin>
FYI I still have that conda build on my list... I just have not gotten back to it :)
< rcurtin>
I might try opening another PR to play around with it, but I'm not sure when I'll have the time
< rcurtin>
the whole thing is way harder than I would have hoped :(
< himanshu_pathak[>
> I might try opening another PR to play around with it, but I'm not sure when I'll have the time
< himanshu_pathak[>
Sure also please look at previous pr before opening new because marcelotrevisani tried to solve the problem and added some commits in my pr those changes were helpful
< himanshu_pathak[>
> the whole thing is way harder than I would have hoped :(
< himanshu_pathak[>
Yeah everytime new problem comes in our path.
< rcurtin>
sounds good---I'll see what marcelo's done and maybe that can be helpful too
< rcurtin>
it may still be some weeks before I get to it though. I just wanted to let you know that it has not (and will not) fall off the list though :)
< himanshu_pathak[>
> it may still be some weeks before I get to it though. I just wanted to let you know that it has not (and will not) fall off the list though :)