rcurtin_irc changed the topic of #mlpack to: mlpack: a scalable machine learning library (https://www.mlpack.org/) -- channel logs: https://libera.irclog.whitequark.org/mlpack -- NOTE: messages sent here might not be seen by bridged users on matrix, gitter, or slack
Guest_li has joined #mlpack
Guest_li has quit [Ping timeout: 244 seconds]
<akhunti1[m]> [microservice/mlpack-3.1.1/build/include/mlpack/prereqs.hpp:74:10: fatal error: boost/serialization/serialization.hpp: No such file or directory
<akhunti1[m]> #include <boost/serialization/serialization.hpp> ]
<akhunti1[m]>
<akhunti1[m]> Hi Team , i am getting this error while compiling mlpack 3.1.1.did i miss some file ?could you please advice
<jonpsy[m]> Eshaan Agarwal: where are we with the PR changes?
<shrit[m]> mlpack 3.1.1 is like decades old. It will simple not work with your settings,
<shrit[m]> Just pull the master one
<rcurtin[m]> agreed, 3.1.1 is very old; in any case, the error indicates that you do not have boost serialization included properly
<EshaanAgarwal[m]> <jonpsy[m]> "Eshaan Agarwal: where are we..." <- Almost done ! Just trying to ease out some errors which are left
<jonpsy[m]> I see, I thought you said its a quick change so I reverted back quickly
<EshaanAgarwal[m]> jonpsy[m]: It was supposed to be quick but as soon as I wrote the test and compiled it. I got a couple errors.
<jonpsy[m]> zoq: would you be able to push your code today?
<jonpsy[m]> EshaanAgarwal[m]: You could've posted them here, anyway , are they resolved?
<EshaanAgarwal[m]> jonpsy[m]: Yeah they are simple but as soon as I am resolving . I am getting an error somewhere else.
<jonpsy[m]> describe them
<jonpsy[m]> whats the error exactly, can u show me a pastebin?
<EshaanAgarwal[m]> jonpsy[m]: Mostly related to declaration and function calls.
<jonpsy[m]> share the pastebin
<EshaanAgarwal[m]> jonpsy[m]: Just a minute
<EshaanAgarwal[m]> EshaanAgarwal[m]: https://pastebin.com/0jgVtvAB
<jonpsy[m]> BitFlipping
<jonpsy[m]> where is this coming from?
<EshaanAgarwal[m]> jonpsy[m]: in HER i am storing all the transitions in a vector
<jonpsy[m]> share me the type of transition
<jonpsy[m]> what exactly is `transition`
<EshaanAgarwal[m]> jonpsy[m]: its a struct of state next state isEnd goal etc
<EshaanAgarwal[m]> i have solved it. I actually passed the arguments in wrong order
<EshaanAgarwal[m]> just got a new one 🫠 Let me look at that one
<jonpsy[m]> haha, send that pastebin then
<jonpsy[m]> basically your constructor was failing
<jonpsy[m]> for `transition`
<EshaanAgarwal[m]> jonpsy[m]: nope i was pushing back the tuple of transition in wrong order. For ex {nextState, state} or {state, nextState}
<EshaanAgarwal[m]> jonpsy[m]: yup one minute
<EshaanAgarwal[m]> EshaanAgarwal[m]: looks like i managed that one too. Phew
<jonpsy[m]> EshaanAgarwal[m]: yes, which is why it couldn't "construct" your object during `push_back`. Since the real ctor expected a different order of inputs
<EshaanAgarwal[m]> EshaanAgarwal[m]: test got compiled successfully finally.
<EshaanAgarwal[m]> jonpsy[m]: yup you are right
<jonpsy[m]> onwards to testing them, do it rn. LEmme know th results
<EshaanAgarwal[m]> jonpsy[m]: yes ! just doing it in a minute. Will keep you posted
<jonpsy[m]> For adding to testsuite, bitflipping is fine. But for actually checking its workin, we need to go for a beter env.
<EshaanAgarwal[m]> jonpsy[m]: sure ! what do you propose for that ?
<EshaanAgarwal[m]> jonpsy[m]: Looks like it failed.
<EshaanAgarwal[m]> due to unexpected exception with message:
<EshaanAgarwal[m]> copy into submatrix: incompatible matrix dimensions: 1x1 and 10x1
<EshaanAgarwal[m]> jonpsy[m]: interesting
<jonpsy[m]> Can't we use our mlpack python bindings, and directly test its working on this repo?
<jonpsy[m]> anyway, print out everything. Where our bitflipping failed
<EshaanAgarwal[m]> jonpsy[m]: I am not sure. Probably zoq can tell us better
<EshaanAgarwal[m]> jonpsy[m]: yup will do it in sometime.
<EshaanAgarwal[m]> should i meanwhile push whatever code i have written till now ?
<jonpsy[m]> ofc
<EshaanAgarwal[m]> cool will do that.
<jonpsy[m]> <EshaanAgarwal[m]> "yup will do it in sometime...." <- Update me on this when you can
<jonpsy[m]> > <@eshaanagarwal:matrix.org> due to unexpected exception with message:
<jonpsy[m]> > copy into submatrix: incompatible matrix dimensions: 1x1 and 10x1
<jonpsy[m]> either env. was setup wrong or something else. Bitflipping, you set 10 as bit length?
<EshaanAgarwal[m]> jonpsy[m]: Default
<jonpsy[m]> and default happens to be 10?
<EshaanAgarwal[m]> jonpsy[m]: Looks like it or maybe HER has some issues with buffer
<jonpsy[m]> it looks like its 1
<jonpsy[m]> do you know how to use GDB?
<EshaanAgarwal[m]> jonpsy[m]: How ? 👀
<EshaanAgarwal[m]> jonpsy[m]: Not much. I have heard about the debugger though
<jonpsy[m]> EshaanAgarwal[m]: How have you been debugging till now?
<EshaanAgarwal[m]> jonpsy[m]: I have set a private variable called length for the length of binary vector
<EshaanAgarwal[m]> jonpsy[m]: Pretty much without that 😅
<jonpsy[m]> how exactly?
<jonpsy[m]> VSC right?
<EshaanAgarwal[m]> jonpsy[m]: Yes
<jonpsy[m]> you do realize VSC uses GDB internally?
<EshaanAgarwal[m]> jonpsy[m]: I did not know
<jonpsy[m]> I mean it just gives you in a prettier format, bt that doens't change the fact
<jonpsy[m]> basially a wrapper around gdb
<EshaanAgarwal[m]> jonpsy[m]: Ok ! I actually did not face much issues related to matrices. In the PPO one we used the C++ program and printed everything
<EshaanAgarwal[m]> Coming back to the issue. I am not sure if I get the issue correctly ? I do understand it's related to the dimension and length
<jonpsy[m]> that error was runtime, i presume?
<EshaanAgarwal[m]> jonpsy[m]: Yeah I ran the test
<jonpsy[m]> ofc, so when its matrix multiplying. The end and start dimension should match
<EshaanAgarwal[m]> jonpsy[m]: Yes !
<jonpsy[m]> not all the test in that file. Just this single test
<EshaanAgarwal[m]> <jonpsy[m]> "> <@jonpsy:matrix.org> sent an..." <- So I change this variable right ?
<jonpsy[m]> can you run `BitFlippingWithNStepHindsightDQN` just this test
<EshaanAgarwal[m]> jonpsy[m]: I ran only this and just gave you the error I got
<jonpsy[m]> yh
<jonpsy[m]> EshaanAgarwal[m]: uh, so it should be 10 ig?
<jonpsy[m]> try replacin with 10 and re-runnin
<EshaanAgarwal[m]> jonpsy[m]: But we are making it a variable right ?
<EshaanAgarwal[m]> jonpsy[m]: Ok ! Give me a minute
<jonpsy[m]> so?
<EshaanAgarwal[m]> jonpsy[m]: It's compiling ! Taking a little bit time.
<EshaanAgarwal[m]> <EshaanAgarwal[m]> "It's compiling ! Taking a little..." <- got segmentation fault
<jonpsy[m]> * oh well 🤦
<jonpsy[m]> oh well 👊
<jonpsy[m]> lets looks at it tmrw
<EshaanAgarwal[m]> yup looks like issue with buffer size. justa a hunch
<EshaanAgarwal[m]> jonpsy[m]: yup ! i will see what i can dig till then
<jonpsy[m]> That's fine, but we should never get `segmentation faults` . Once you find your error, put appropriate `asserts` so that its caught in compile time
<EshaanAgarwal[m]> jonpsy[m]: understood
<jonpsy[m]> can't stress the "never" enough. It creates bad image
<EshaanAgarwal[m]> jonpsy[m]: i understand 😢
<EshaanAgarwal[m]> will look in it then ! will post if i get something
<akhunti1[m]> <rcurtin[m]> "agreed, 3.1.1 is very old; in..." <- cmake_minimum_required(VERSION 3.4.1)... (full message at <https://libera.ems.host/_matrix/media/r0/download/libera.chat/64976b70f77ce15ad438b6820c513b34b8046070>)
<akhunti1[m]> Hi rcurtin this is my cmakelist.txt file . i added boost library inside my project folder. but still i am getting [ Could NOT find Boost (missing: Boost_INCLUDE_DIR serialization] this error.
<akhunti1[m]> As our project already build in mlpack 3.1.1., so it is very difficult for us to go for newer version of mlpack.
<rcurtin[m]> I think most things are reverse compatible from 3.4.2 to 3.1.1, so it may be easier than you think to upgrade, but okay
<rcurtin[m]> if CMake is not finding Boost successfully, make sure it's installed on the machine, and if you installed it to a custom directory you might try specifying the `BOOST_ROOT` option to CMake