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/
dcharrezt has quit [Ping timeout: 240 seconds]
< Param-29Gitter[m>
Hello , I tried updating naive_method.hpp and tested on CircleTransformationTestNaive with number of samples = 1200. #2286
< PrinceGuptaGitte>
Thanks a lot for the suggestions @zoq
ImQ009 has joined #mlpack
favre49 has joined #mlpack
favre49 has quit [Ping timeout: 240 seconds]
favre49 has joined #mlpack
< favre49>
I've been getting weird behavior make-ing mlpack
< favre49>
If I run with multiple processes (make -j4) I quickly run into a situation where all the processes enter disk sleep and use all my memory
< favre49>
The processes just stay in disk sleep, as if it's some kind of deadlock
< favre49>
But if I run it with just a single process, I have no problems. Any idea why that might be?
< AbishaiEbenezerG>
oh wow
< AbishaiEbenezerG>
i'm having the same problem right here
< AbishaiEbenezerG>
when i did the make with just a single process, it took a really long time, but not too many problems
< favre49>
Could you tell me your system specs?
< AbishaiEbenezerG>
now, after i cloned mlpack from github and i decided to build it in a different location, i decided to make it with -j4
< AbishaiEbenezerG>
and its taking forever...
< favre49>
Could you use htop or some other resource monitor to see the status of your processes?
< AbishaiEbenezerG>
8 gb ram and i7-8th gen
< AbishaiEbenezerG>
i have no clue why this is happening. i didn't even go to college today so that i could work on mlpack locally, and this build is taking all day
< favre49>
That's some dedication :) Perhaps it's a simple issue though, I don't know enough to diagnose it myself. Hopefully someone else knows what the issue is
< AbishaiEbenezerG>
sure,i'll try looking at my resource monitor....but the hang is so bad, that if i try opening an application now, it will open like 4 min later...it seriously is very frustrating
< favre49>
I'd advise just restarting your laptop. That's what i did, when i faced this issue
< AbishaiEbenezerG>
here's my story- i started the build at 11 in the morning. At around 1, it got till about 75% and then it was practically dead. Then, i restarted my laptop and started all over again
< AbishaiEbenezerG>
and here i am now
< AbishaiEbenezerG>
and its at 78% currently
< AbishaiEbenezerG>
looks like i'm going to have a long night, cuz i basically haven't even gotten close to what i set out to do today. very very frustrating. i'm so sorry for putting my pain out here, but i really wish i knew how to fix this up
< PrinceGuptaGitte>
Even with just 1 core make-ing mlpack , it shouldn't take this long.
< PrinceGuptaGitte>
considering your system specs are pretty good
< zoq>
AbishaiEbenezerG: What was the last commit for the other build that worked?
< AbishaiEbenezerG>
yeah, with 1 core not much problem there
favre49 has quit [Remote host closed the connection]
< AbishaiEbenezerG>
@zoq so that build was on 23rd feb 8:30 pm ist
< AbishaiEbenezerG>
the build is finally over and things are back to normal ig. but yeah, this something should check out
< zoq>
AbishaiEbenezerG: can you post the commit hash?
< AbishaiEbenezerG>
sure.
< AbishaiEbenezerG>
i'm on the network page on github
< AbishaiEbenezerG>
how do i see the commits with respect to dates?
< zoq>
If you still have the repo you run "git log" to see the commits
< AbishaiEbenezerG>
ok so looks like #2206 and #2219 were the 2 previous ones...
< AbishaiEbenezerG>
not sure how much this would help, but at that time , i followed the exact instructions as given in the documentation and did it using just a single core
< AbishaiEbenezerG>
today, i had cloned it from the github repo and started of with cmake ../ and then make -j4
< AbishaiEbenezerG>
*off
< zoq>
My guess is since you buld with j4 it fills up all your memory, and starts to swap out to the disk.
< AbishaiEbenezerG>
another question
< AbishaiEbenezerG>
the size of the 23rd feb repo is 8.9 gb whereas the one i got today is just 3.4 gb. why the difference?
< zoq>
I guess including the build files?
< AbishaiEbenezerG>
yes. both are including build files...
< zoq>
Not sure, maybe you build the other one with debug symbols?
< zoq>
or you have some other data in there
< zoq>
just build with -j4 without any problems on my system
zalava[m] has joined #mlpack
< AbishaiEbenezerG>
actually there are more items in todays repo- 100 more, and it still is smaller.
< AbishaiEbenezerG>
idk, maybe something wrong on my side
< zoq>
you would have to make sure you used the same build settings for both versions
< AbishaiEbenezerG>
i thought something might be wrong as even @favre49 had the same problem...
< AbishaiEbenezerG>
could the difference in size be becuz i had downloaded the zip and then did a tar and then did the make...?
< AbishaiEbenezerG>
this was for the earlier version
< zoq>
I thought you used the git master branch?
< AbishaiEbenezerG>
oh no
< AbishaiEbenezerG>
i used the master today
< AbishaiEbenezerG>
i was a total noob and so i had just followed the documentation blindly on 23rd
< AbishaiEbenezerG>
thats when i did it with just a single core ...
< zoq>
Hm, but you last change was #2206? I'm confused...
< AbishaiEbenezerG>
@zoq i think i'm getting it all messed up here, so let me clarify
< AbishaiEbenezerG>
i don't think there's any correlation between my 23rd build and #2206
< AbishaiEbenezerG>
i used the master branch today
< AbishaiEbenezerG>
the build on 23rd was what i followed from the docs i.e the download of the zip and then tar that and so on...
< zoq>
AbishaiEbenezerG: okay, so if you build the master branch with multiple cores, did you track the memory usage?
< AbishaiEbenezerG>
i'm sorry if i'm really confusing you. but i think if the j4 worked fine on your system, then there shouldn't be any problems ig. Most probably, i've got something wrong here, or something wrong in my understanding...
< AbishaiEbenezerG>
nope , i didn't . the system was like in a semi-hung state, so i really didn't bother trying anything. I thought i'll just leave it alone....i should have checked it tho
< AbishaiEbenezerG>
I'm pretty new to open source contribution, so i still some time and space..
< AbishaiEbenezerG>
i have one more question @zoq- now that i have my own local version of mlpack, when i make changes to the source code, doo i need to build mlpack again?
< zoq>
No problem at all, it's always difficult to jump into a new project.
< zoq>
yes, but it will only build the changes
< AbishaiEbenezerG>
awesome
< AbishaiEbenezerG>
thanks @zoq. I'm doing what i can to get better. Thank you so much for the assistance!
< zoq>
Happy to help.
< pickle-rick[m]>
Has anyone seen this sort of error when trying to compile a cpp file that uses ffn?
< pickle-rick[m]>
/usr/include/boost/serialization/access.hpp:116:11: error: ‘class arma::Mat<double>’ has no member named ‘serialize’; did you mean ‘set_size’?
< pickle-rick[m]>
t.serialize(ar, file_version);
< rcurtin>
pickle-rick[m]: this happens if you do #include <armadillo> before #include <mlpack/core.hpp>
< rcurtin>
the reason is that mlpack/core.hpp (or really mlpack/core/arma_extend/arma_extend.hpp) includes Armadillo in a special way that defines a serialize() function for Armadillo types
< rcurtin>
but, it doesn't work if you do #include <armadillo> first :)
< rcurtin>
in fact, in this case, the compiler probably printed a big warning about including Armadillo first :)
< pickle-rick[m]>
Yeah the compiler did :), and I thought that would be a separate issue to deal with. But now I know it's not. I am yet to see where I've included armadillo, but let me go through all the code again. Thanks for this info.
< rcurtin>
I didn't like that that's what we had to do to get serialization support in Armadillo---it's not fun to force include orderings on users
< rcurtin>
but I haven't seen any other way to do it. maybe it's possible in some other way, but I haven't looked into it in quite some time... it was one of those things where I got it working and said "good enough" and moved on :)
< pickle-rick[m]>
ah. I see. It is what it is right 😇
< rcurtin>
yeah :)
MrityunjayTripat has joined #mlpack
< MrityunjayTripat>
Slack messages are not sent :(
< MrityunjayTripat>
Or rather not seen on gutter?
< MrityunjayTripat>
*gitter
< rcurtin>
MrityunjayTripat: hmmm, let me check on it
< rcurtin>
MrityunjayTripat: I guess that's why none of us have been responding to your messages :-D
letrunghieu[m] has left #mlpack []
BisakhMondal[m] has left #mlpack []
ZishanSheikh[m] has left #mlpack []
zoq[m] has left #mlpack []
M_slack_9 has left #mlpack []
EdmundWu[m] has left #mlpack []
ac-optimus[m] has left #mlpack []
M_slack_mlpack_4 has left #mlpack []
M_slack_mlpack10 has left #mlpack []
RohitKBharadwaj[ has left #mlpack []
sailor[m] has left #mlpack []
Shikhar[m] has left #mlpack []
KimSangYeon-DGU4 has left #mlpack []
HimanshuPathak[m has left #mlpack []
Randal[m] has left #mlpack []
chopper_inbound[ has left #mlpack []
Hunting-Party10[ has left #mlpack []
ToshalAgrawal[m] has left #mlpack []
Nipunwaas[m] has left #mlpack []
Saksham[m] has left #mlpack []
birm[m] has left #mlpack []
JaskaranKalra[m] has left #mlpack []
ryan[m] has left #mlpack []
VikramSinghChund has left #mlpack []
M_slack_mlpack_U has left #mlpack []
Keyur[m] has left #mlpack []
Srishti[m] has left #mlpack []
M_slack_mlpack_7 has left #mlpack []
ArunavShandeelya has left #mlpack []
SlackIntegration has left #mlpack []
SlackIntegration has joined #mlpack
ryan[m] has joined #mlpack
< ryan[m]>
test message after relinking rooms
< ryan[m]>
chopper_inbound okay, that seemed to work. I have no idea why... I just disabled the bridge and reenabled it
chopper_inbound[ has joined #mlpack
< rcurtin>
I don't often have mlpack slack open, so if the problem happens again speak up and I can do the same thing :)
< SriramSKGitter[m>
This might seem like a silly question, but where are the .html files created by doxygen stored after building?
< pickle-rick[m]>
Its in build/doc/html
< SriramSKGitter[m>
@pickle-rick : Thanks for the quick reply. However, I don't appear to have a doc/ directory in build/
< pickle-rick[m]>
ah i see
< pickle-rick[m]>
you'd have to do a "make doc" for that
< pickle-rick[m]>
after going to your build directory
< SriramSKGitter[m>
Doesn't `make` default to `make all` which should include `make doc`?
< rcurtin>
SriramSKGitter[m: it does, but I don't think the `doc` target is available if you don't have doxygen installed on your system
< rcurtin>
but if `make doc` works but `make` doesn't do `make doc`, it sounds like an issue... if you want to either fix it by modifying CMakeLists.txt and submit a PR, or open an issue about it, I think that would be a good way forward :)
< SriramSKGitter[m>
@rcurtin : It is installed, shows up when I run `cmake ../`, although it does show `found components: doxygen missing components: dot`
< SakshamRastogiGi>
Can confrim this
< SakshamRastogiGi>
doc folder was not visible after make but after make doc inside build it shows up
< SriramSKGitter[m>
I looked around and the missing component dot doesn't seem to be of any consequence
< SriramSKGitter[m>
@codeboy5 : Looks like CmakeLists.txt will have to be modified then
< pickle-rick[m]>
<rcurtin "but if `make doc` works but `mak"> ohh, i thought it was something make didn't do by default, because I already had doxygen installed as well.
< SakshamRastogiGi>
I thought it might be choice, in case someone needs the documentation they can build it
< pickle-rick[m]>
<SakshamRastogiGi "I thought it might be choice, in"> same.
< SriramSKGitter[m>
Guess I'm the only one who thinks otherwise :) I would expect `make` to generate the docs by default
< nishantkr18[m]>
Hey!
< nishantkr18[m]>
I am actually stuck in a problem, I try this for running without installing `g++ -g /home/nishantkr18/Desktop/TEST/test.cpp -o /home/nishantkr18/Desktop/TEST/test -std=c++11 -fopenmp -I/home/nishantkr18/Desktop/mlpack-3.2.2/build/include/ -L/home/nishantkr18/Desktop/mlpack-3.2.2/build/lib/ -larmadillo -lboost_serialization -lboost_program_options`
< nishantkr18[m]>
<nishantkr18[m] "I am actually stuck in a problem"> its giving me ` undefined reference to `mlpack::util::GetVersion[abi:cxx11]()' ` as error
< nishantkr18[m]>
<nishantkr18[m] "I am actually stuck in a problem"> Is the correct way to run without installing?
< nishantkr18[m]>
> <@nishantkr18:matrix.org> I am actually stuck in a problem, I try this for running without installing `g++ -g /home/nishantkr18/Desktop/TEST/test.cpp -o /home/nishantkr18/Desktop/TEST/test -std=c++11 -fopenmp -I/home/nishantkr18/Desktop/mlpack-3.2.2/build/include/ -L/home/nishantkr18/Desktop/mlpack-3.2.2/build/lib/ -larmadillo -lboost_serialization -lboost_program_options`
< nishantkr18[m]>
* Is this the correct way to run without installing?
< zoq>
nishantkr18[m]: It's somewhat complex to build only a single file, so I would build mlpack first and link against the lib afterwards.
< zoq>
SriramSKGitter[m: I think most people building an unstable version (e.g. git master branch) aren't interested in the doc.
< chopper_inbound[>
ryan: Ok. Let me take a deeper look then...
< SriramSKGitter[m>
@zoq : I guess you're right. I'll just compile them manually when I need to have a look :)
< rcurtin>
nishantkr18[m]: it looks like your compilation command there isn't actually linking against mlpack and needs -lmlpack
< rcurtin>
zoq: fair point, maybe it's better leave `doc` out of `all`
< rcurtin>
so, our CMake configuration links against boost_unit_test_framework.so, which suggests that boost/test/unit_test.hpp is the right one to include
< rcurtin>
if you're trying to compile *_test.cpp files by hand, I'd suggest not doing that; much better to use CMake+make to build mlpack_test
< chopper_inbound4>
rcurtin: Got that! Thanks for the help :)
< nishantkr18[m]>
<zoq "nishantkr18: It's somewhat compl"> I had already built mlpack, and linked with `-I/home/nishantkr18/Desktop/mlpack-3.2.2/build/include/ -L/home/nishantkr18/Desktop/mlpack-3.2.2/build/lib/ `
< nishantkr18[m]>
<rcurtin "nishantkr18: it looks like your "> but for -lmlpack, I need to install mlpack right?
< nishantkr18[m]>
for development purposes, should one always install and use? would it not be better to just build the changes without installing?
< kritika12298Gitt>
How to get started with Armadillo, actually I installed it but still i'm not getting how to use it with Dev-C++ on Windows
< rcurtin>
nishantkr18[m]: nah, no need to install for just -lmlpack; with the -L option, you are specifying the directory that gcc should look in for libraries, and with -lmlpack, you tell it to look for libmlpack.so in whatever directories it knows to look for libraries in
< rcurtin>
and since you added -L/home/nishantkr18/Desktop/mlpack-3.2.2/build/lib/, it will search in there for libmlpack.so and everything should link correctly :)
< nishantkr18[m]>
<rcurtin "and since you added -L/home/nish"> I too thought the same, not working though.. I'll keep trying , Maybe i m missing something
< rcurtin>
does libmlpack.so exist in /home/nishantkr18/Desktop/mlpack-3.2.2/build/lib/?
< nishantkr18[m]>
<rcurtin "does libmlpack.so exist in /home"> yes, it does
< rcurtin>
okay, strange, in this case I am not sure, maybe there are some other libraries that need to be linked too