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/
ImQ009 has joined #mlpack
travis-ci has joined #mlpack
< travis-ci>
UtR491/ensmallen#24 (moead_exp - 6b4239f : UtR491): The build was broken.
< shrit[m]>
That is good to know that I need at least potato on my machine
< shrit[m]>
I will try to do the installation today
< jeffin143[m]>
> jeffin143: zoq: I use mutt for email so I don't see images unless I open the link... I am looking through some of these catch PR gifs and laughing really hard
< jeffin143[m]>
@rcurtin:matrix.org indeed , it was zoq idea to spice it up and I got so much hit by it that now every pr of mine has one of those
< jeffin143[m]>
Even brim joined us yesterday 😂
< shrit[m]>
rcurtin: The CD images of debian archive seems to be working perfectly. Knowing that I have never installed anything as old as this potato
travis-ci has joined #mlpack
< travis-ci>
UtR491/ensmallen#25 (moead_exp - 8ba5cec : UtR491): The build was fixed.
< shrit[m]>
Funny enough that you can boot from a CD but you need a floppy disk to install the kernel since the potato installer does not recognize the cdrom
< shrit[m]>
Finally, I was able to install debian woody using network installer
Eddie56 has joined #mlpack
< Eddie56>
Have any of you guys used RandomForestRegressor of MLPack?
Eddie56 has quit [Ping timeout: 245 seconds]
< rcurtin>
shrit[m]: nice! what are the specs on that laptop?
< rcurtin>
my 386 is still upgrading to potato... ~12 hours now
< shrit[m]>
CPU : Pentium 3 max 694MHZ
< shrit[m]>
64 MB RAM
< shrit[m]>
16 GB HDD
< shrit[m]>
ATI graphic card
< rcurtin>
nice, almost modern :-D
< shrit[m]>
12 inch screen
< shrit[m]>
modem + ethernet
< shrit[m]>
+ WiFi as extension card
< rcurtin>
PCMCIA?
< rcurtin>
(for the wifi card)
< rcurtin>
I used to have a Thinkpad 390X; it was my first computer. I remember the day I bought a PCMCIA wifi card for it... it suddenly became like 1000x more useful
< rcurtin>
that was back when nobody ever even secured their wifi at all (not even WEP), so it was like there was free internet everywhere (or it seemed like it)
< shrit[m]>
Yes PCMCIA
< rcurtin>
:)
< shrit[m]>
The issue is that the kernel does not recongnize the PCMCIA
< shrit[m]>
and also there is no ssh in debian woody
< shrit[m]>
which is very wierd
< shrit[m]>
There should be a video chat today right?
< rcurtin>
you might need to make sure the contrib and non-free repos are enabled too
< shrit[m]>
OK, because it is telling me that there is no available version, but exists in the database
< rcurtin>
strange
< shrit[m]>
I have hamm contrib main non free
< shrit[m]>
Yeap
< rcurtin>
hamm? shouldn't that be woody?
< rcurtin>
hamm is actually older than slink
< shrit[m]>
Oh, I wrote that manually,
< shrit[m]>
Hope that I did not get a wrong sources
< shrit[m]>
I had to add source manually.
< rcurtin>
yeah, same here, I bet if you change that to woody it will fix some things
< rcurtin>
it might need to upgrade lots of packages though :)
< shrit[m]>
Well I am installing everything one by one
< rcurtin>
ouch, time-consuming :)
< shrit[m]>
I basicallty spent the day on the installation, even for woody it was impossible to install from CDROM, since the installer does not recongnaize the driver. Finally I was able to install using network installer
< rcurtin>
yeah, for my 386 it was basically the same story. it was super time-consuming to figure out the right combination of things that let me actually get through the install :)
< rcurtin>
I forget how hard Linux used to be to install
< shrit[m]>
Exactly
< shrit[m]>
Old debian is harder than modern day arch
< rcurtin>
although these days, I think it is getting hard again. but this could be because I am just not knowledgeable about UEFI and all of the new strange boot and security technologies
< shrit[m]>
I got ssh by changing it to woody
< shrit[m]>
I installed something called git. It seems to be called Gnu Interactive Tool
< rcurtin>
ha!
< rcurtin>
yeah, I imagine you will have to compile git from source, or just, e.g., `wget mlpack.org/files/mlpack-3.4.0.tar.gz' and then install like that
< rcurtin>
but even that will be hard
< rcurtin>
getting CMake 3 on that system will probably be super hard
< rcurtin>
not to mention a new enough gcc
< rcurtin>
for the 386 I was planning to just cross-compile... anything else would be impossible with only 8MB of RAM
< rcurtin>
given that the mlpack tests usually take several GB of RAM to compile :-D
< shrit[m]>
I think I am going to compile the binaries and use scp
< rcurtin>
probably a bit easier like that :)
< rcurtin>
with a pentium 3... you *might* be able to get a new enough debian version on it
< rcurtin>
I think that you *could* even get the current stable version running; kernel support for the P3 should still be there
< rcurtin>
but having only 64MB RAM could pose a problem, not sure
< shrit[m]>
Yes, I have already consumed 50 MB
< shrit[m]>
58 to be precise
< shrit[m]>
58 out of 64
< rcurtin>
ha
< rcurtin>
whew
< shrit[m]>
Do you have any idea how to install the xserver ? it is missing configuration
< rcurtin>
it might be called XFree86 or something like this, and you'll probably have to manually write an xorg.conf
< rcurtin>
I remember doing this with some of the first linux machines I worked on, and it was basically awful and always broke
< rcurtin>
if it's an ATI video card, you might also need the 'fglrx' kernel module, and I remember that also being a bit of a nightmare
< rcurtin>
I think you can use the 'vesa' driver, but you won't get direct rendering and it will be slow (maybe that doesn't matter if you aren't using OpenGL or anything)
< abernauer[m]>
Have to set up my first phone screen out of my latest batch of 140 applications.
< rcurtin>
abernauer[m]: awesome, hope it goes well :)
< abernauer[m]>
rcurtin: Thanks. Yeah it's kind of difficult to find good opportunities in Data Science where there is a right intersection with my skillset, experience, and being able to use the tech I actually enjoy.
< AakashkaushikGit>
hey, I had a doubt that is it okay if i use arma::mean((input-target)/input) instead of first summing ((input-target)/input) and then dividing by rows to write the mean absolute error loss, Does it have ant performance cost or is there any better way that i should use?
< AakashkaushikGit>
(edited) ... have ant performance ... => ... have any performance ...
mlpack-meeting has quit [Remote host closed the connection]
< anjishnu[m]>
"Aakash kaushik (Gitter)" (https://matrix.to/#/@gitter_aakash-kaushik:matrix.org) You would need to be able to return both the sum reduction of the loss function as well as the mean reduction via use of the reduction parameter in the constructor. So, first calculating the sum and then the mean by division makes more sense I guess.
< AakashkaushikGit>
Sure, it does makes sense, will take the latter approach then.
travis-ci has joined #mlpack
< travis-ci>
UtR491/ensmallen#27 (moead - c69171c : UtR491): The build was broken.
< AakashkaushikGit>
Hi, I have written the MAE function and also added it to the cmake file in methods/ann/loss_function, how should i test i now so i can create a PR ?
< anjishnu[m]>
Write a test case for it in src/mlpack/tests/loss_functions_test.cpp, then compile the test cases using "make mlpack_catch_test" and then run that particular test case using "./bin/mlpack_catch_test "name_of_test_case""
< AakashkaushikGit>
also i am never able to make the whole mlpack, i always get this error at around 50%
< AakashkaushikGit>
I feel like this can be because of a ram issue and if that's so is there any way to build it ?
< rcurtin>
AakashkaushikGit: exactly, you are out of RAM... are you building with only one core?
< rcurtin>
like 'make mlpack_catch_test' instead of 'make -j2 mlpack_catch_test'
< rcurtin>
unfortunately mlpack is pretty mean with RAM usage during compilation...
< anjishnu[m]>
I just remembered the -j8 I use but forgot to mention above :)
< AakashkaushikGit>
i am building it with a single core but i have a i5-8265U with 4 cores and i did try with 2 and 4 cores it always fails
< abernauer[m]>
rcurtin: The R bindings are not on CRAN at this point correct?
< rcurtin>
AakashkaushikGit: try with just 1 core?
< rcurtin>
abernauer[m]: right, not yet
< AakashkaushikGit>
@rcurtin trying right now
< AakashkaushikGit>
Finally done had to close everything else and with a single core.
< rcurtin>
AakashkaushikGit: awesome, glad it worked... sorry it takes so much RAM :(
< zoq>
AakashkaushikGit: Maybe you can increase SWAP as well, so you don't have to close everything else.
< AakashkaushikGit>
I had that thought in mind and was going that route but then i read somewhere that it is not advisable on SSDs so i didn't went further with it, also i did check that i have a 2gb swap already created.
< AakashkaushikGit>
Not advisable because of the constant reads and writes i think.