mlpack IRC logs, 2020-03-10

Logs for the day 2020-03-10 (starts at 0:00 UTC) are shown below.

March 2020
--- Log opened Tue Mar 10 00:00:22 2020
03:34 < kartikdutt18Gitt> Thank you very much @zoq . I will refine my proposal a bit according to the changes you suggested and get back to you. Thanks.
06:45 -!- louisway [] has joined #mlpack
06:51 -!- louisway [] has quit [Remote host closed the connection]
07:05 < jenkins-mlpack2> Project docker mlpack nightly build build #636: STILL FAILING in 2 hr 51 min:
11:39 -!- togo [~togo@2a02:6d40:34db:fd01:64ff:b1f7:4e47:228e] has joined #mlpack
11:49 < pickle-rick[m]> Hey there, is there a way to make bindings for python3 instead of python2 when running cmake? It always discovers python2 when I run cmake.
11:52 < zoq> pickle-rick[m]: Yes, add -DPYTHON_EXECUTABLE=/usr/bin/python to the cmake step.
11:52 < zoq> kartikdutt18Gitt: Sounds good.
11:54 < pickle-rick[m]> <zoq "pickle-rick: Yes, add -DPYTHON_E"> Oh nice. Thanks
12:26 < SriramSKGitter[m> Has anyone used for OSX VMs? Is this reliable? Or does anyone know of an alternate solution?
12:27 < pickle-rick[m]> Are there dependencies apart from python3-numpy, python3-pandas and python3-setuptools, that I need to install to successfully build mlpack with python3.
12:28 < pickle-rick[m]> I guess the dependencies given on the mlpack website covers dependencies only for python2? Am I correct in assuming that?
12:28 < zoq> pickle-rick[m]: cython setuptools distutils numpy pandas
12:29 < zoq> pickle-rick[m]: same for python3 and python2
12:29 < pickle-rick[m]> I see
12:30 < zoq> SriramSKGitter[m: Never used that part, we used/use travis and azure to run tests against macOS
12:33 < SriramSKGitter[m> @zoq : The macOS tests on #2241 are failing with an unhelpful error message. So I'm trying to get my hands on an OSX VM to build from scratch and find out what's wrong
12:34 < SriramSKGitter[m> I could use my brother's Macbook, but it's a company issued laptop and I'd rather not mess with it :)
12:35 < zoq> SriramSKGitter[m: you mean the hdf5 error?
12:35 < zoq> SriramSKGitter[m: I think this is a package issue.
12:36 < zoq> SriramSKGitter[m: The other issue can be replicated on a Linux machine as well, so no need to look for a macOS machine.
12:36 < zoq> SriramSKGitter[m: and we have some members that have access to a mac machine.
12:38 < SriramSKGitter[m> @zoq : I'm not sure about the hdf5 error you've mentioned. The error is a simple unit test I've written for the CosineTree. The test only fails on the macOS builds
12:38 < SriramSKGitter[m> If anyone with a Mac would like to help me with #2241, I'd be grateful :)
12:39 -!- ankur-banga [31248f2c@] has joined #mlpack
12:42 < zoq> SriramSKGitter[m: CosineTreeTest fails on the Linux Plain build as well.
12:42 < zoq> SriramSKGitter[m: Make sure you build with -DDEBUG=ON, to enable bound checking.
12:44 < SriramSKGitter[m> @zoq : Ah, I thought the Linux build failed because of #2257 , but yeah looks like there's something wrong there as well.
12:44 < SriramSKGitter[m> I'll try my luck with the `-DDEBUG=ON` switch
12:47 < kartikdutt18Gitt> Hi @sriramsk1999, The linux build was fixed in #2257, For Mac OS, Some update in a dependency is causing the failure. I tried on my local machine it worked fine, however then I updated my dependencies, then same the load save test failed.
12:48 < kartikdutt18Gitt> I have opened #2271 to investigate the same. Hopefully it will be fixed too.
12:51 < rcurtin> zoq: you're right! thanks for pointing out that CosineTreeTest also failed on Linux
12:51 < rcurtin> I'm not sure how I overlooked the error when I looked through that log originally, sorry about that SriramSKGitter[m
12:53 < SriramSKGitter[m> @kartikdutt18 : Hmm, that's very curious, there was a loadsave test fail on one of my commits as well, but that was resolved after @rcurtin restarted the build.
12:54 < SriramSKGitter[m> What leaves me stumped right now is how my one new test is failing when it doesn't bring any new functionality to the table as far as I can see
12:54 < SriramSKGitter[m> No problem @rcurtin, I overlooked it as well :)
12:59 -!- ankur-banga [31248f2c@] has quit [Remote host closed the connection]
13:02 < kartikdutt18Gitt> Well all builds that were made yesterday failed LoadSaveTest. And I think the same goes for pretty much every build today. (I have also made the error occur on my local machine so I am thinking this is not a random error but something that's needed to be fixed, I might be missing something though.)
13:16 < AbishaiEbenezerG> got a question here- having experience of rl to a certain level, what should i do from here after having gone through the codebase? Need a bit of guidance here...
13:17 < PrinceGuptaGitte> You can try replicating some basic example using MLPack. It's always a good idea
13:17 < AbishaiEbenezerG> ok
13:18 < AbishaiEbenezerG> you mean i could try implementing something i have learnt with mlpack ..right?
13:18 < AbishaiEbenezerG> sorry if that didn't make sense
13:19 < PrinceGuptaGitte> yes
13:19 < PrinceGuptaGitte> like hello world of reinforced learning
13:19 < AbishaiEbenezerG> got it. Thanks alot !
13:19 < AbishaiEbenezerG> hey btw
13:21 < AbishaiEbenezerG> till now, and probably for a few more days, the number of PRs of mine on github might remain zero
13:21 < AbishaiEbenezerG> would that affect my chances...? i will contribute as much as i can asap...
13:23 < AbishaiEbenezerG> how big a factor is it? I should be in a position to start experimenting with the codebase by thursday/friday. Would it hamper my chances?
13:24 < PrinceGuptaGitte> I don't think it should affect if you contribute decently, but only a mentor can answer it.
13:24 < AbishaiEbenezerG> alright. thanks tho!
13:26 < AbishaiEbenezerG> @zoq any specific advice for me given where i stand right now?
13:29 < AbishaiEbenezerG> or for that matter, any mentor's advice would be great to have
13:34 < rcurtin> SriramSKGitter[m: when I debug serialization tests, one of the things I'll do is comment out the code that removes the temporary file at the end of the serialization tests so that I can `cat` that file and see if anything is wrong
13:34 < rcurtin> I'll also try and determine if the issue is with the XML, text, or binary archives
13:34 < rcurtin> it's usually easier to debug if the issue is XML, since the file is readable
13:35 < rcurtin> and if you can reproduce locally, you can use a debugger like gdb to find the actual line of code that's causing the problem
13:35 < rcurtin> (warning: boost serialization backtraces are big! :))
13:53 < SriramSKGitter[m> @rcurtin : I'm setting up mlpack on a Mac right now, hopefully I can reproduce the error right here and find out the cause. I'll keep all the other methods in mind in case the test runs successfully here as well.
13:54 < rcurtin> sounds good; looks like it would be pretty straightforward to reproduce on Linux too
13:59 < SriramSKGitter[m> > `rcurtin on Freenode` Sriram S K (Gitter): when I debug serialization tests, one of the things I'll do is comment out the code that removes the temporary file at the end of the serialization tests so that I can `cat` that file and see if anything is wrong
13:59 < SriramSKGitter[m> I just had a look at `serialization.hpp`. Is `remove(fileName.c_str());` the line you're talking about?
14:10 < rcurtin> yeah, exactly, if you comment that out you'll get a bunch of temporary files in your build directory after you run the tests
14:10 < rcurtin> I find the most recent one with `ls -lth | head -1` and then cat it
14:10 < rcurtin> (it'll be whatever model got serialized, but failed to deserialize, I think)
14:15 < zoq> AbishaiEbenezerG: Running some experiments is a good idea, you can take a look at the test cases to see some simple examples.
14:16 < SriramSKGitter[m> @rcurtin : Thanks for the help :) I have an arsenal of things to try out :)
14:19 < zoq> AbishaiEbenezerG: Also, the number of PR's doesn't impact your chances, it's a nice way to show you are familiar with the codebase, also we look for quality and not for quantity.
14:24 < zoq> And I think at this point, we can't keep up witht he number of PR's, a review takes time; like if someone implements a new activation function, we have to read the paper, check the implementation and tests.
14:24 < zoq> Not something we can do in like 20 minutes or so.
14:27 < AbishaiEbenezerG> thank you so much @zoq. Very heartening. I'll do some study on the test cases and my own cases and take it from there...
14:28 < zoq> AbishaiEbenezerG: Sounds good :)
15:04 -!- kunal12298Gitter [gitterku_1@gateway/shell/] has joined #mlpack
15:04 < kunal12298Gitter> Please guide me on how to get started
15:04 < kunal12298Gitter> Hello, I am Kunal Gupta and am looking forward to contribute to the project Essential Deep Learning Modules.
15:05 < PrinceGuptaGitte> Hi @kunal12298 have a look at and
15:07 < SriramSKGitter[m> I have a question for the mentors. It
15:08 < SriramSKGitter[m> It's mentioned in the application guide that the project proposal is the most important part of the application. With the large number of ideas on the mailing list, I find many ideas quite similar to my own :) In that case, how would be the proposals be evaluated?
15:08 < kunal12298Gitter> And for the other project, I want to apply for Algorithm Optimization. How should I contact the mentors regarding this project as there are no PRs mentioned with its description
15:15 < saksham189Gitter> @kunal12298 you can ask your questions here
15:17 < kunal12298Gitter> Okay
15:29 -!- ImQ009 [~ImQ009@unaffiliated/imq009] has joined #mlpack
15:56 -!- robertohueso [robertohue@gateway/shell/] has joined #mlpack
15:56 < PrinceGuptaGitte> Hi @zoq , I've added test cases in PR #2243 , but I have a doubt the way layer_names is set up, please have a look when you get time. Thanks
16:02 < jenkins-mlpack2> Yippee, build fixed!
16:02 < jenkins-mlpack2> Project mlpack - git commit test build #338: FIXED in 1 hr 0 min:
16:56 < kartikdutt18Gitt> Hey @rcurtin, Would it be possible to add #2265 to mlpack 3.3.0, it's a bug fix for pooling layers. Currently if we use pooling layers with unequal strides, the pooling layers might throw an out of index bounds error.
17:24 < rcurtin> kartikdutt18Gitt: ah, good point, it looks like a pretty simple change so I think we can add it
17:25 < rcurtin> I'll try and review it later today
17:53 < kartikdutt18Gitt> Great, Thanks a lot.
18:17 -!- snig [] has joined #mlpack
18:47 -!- snig [] has quit [Remote host closed the connection]
18:58 * pickle-rick[m] sent a long message: < >
19:18 < pickle-rick[m]> Also it'd be really helpful, if someone could point me towards documentation on the usage of accuracy metrics. I'm doing something wrong when trying to use accuracy.hpp, don't know what. Thanks in advance.
19:28 < rcurtin> pickle-rick[m]: I thought we had one-hot encoding utilities but they don't appear to be there... I've written some code to do it, though, let me see if I can dig it up to submit as a PR...
19:33 < pickle-rick[m]> Oh wait, is it this one?
19:34 < pickle-rick[m]>
19:44 -!- ImQ009 [~ImQ009@unaffiliated/imq009] has quit [Quit: Leaving]
19:55 -!- louisway [] has joined #mlpack
20:09 -!- ykasture [] has joined #mlpack
20:18 < PrinceGuptaGitte> Hi, I'm trying to deserialize the layers in the `network` of data type `std::vector<LayerTypes<CustomLayers...> > `. But it's not working. Anyone have any idea
20:34 < zoq> PrinceGuptaGitte: Not sure, we can provide much help, because "it's not working" isn't really helpful; maybe you can provide some more context, maybe even a simple example, or the error you get.
20:44 -!- ykasture [] has quit [Remote host closed the connection]
21:47 < PrinceGuptaGitte> I'm really sorry for being so vague about the problem. Actually I was getting an error like: , So I was thinkinh if I'm doing some common mistake. Turns out I wasn't including `-lboost_serialization -lboost_program_options` these flags while compiling
21:47 < PrinceGuptaGitte> (edited) ... sorry for being so vague about the problem. Actually I was getting an error like: , So I was thinkinh if ... => ... sorry @zoq for being so vague about the problem. Actually I was getting an error like: , So I was thinking if ...
22:00 < rcurtin> pickle-rick[m]: ah, there it is! I think what I wrote was a preprocess_onehot_main.cpp (a binding for it)
22:22 -!- togo [~togo@2a02:6d40:34db:fd01:64ff:b1f7:4e47:228e] has quit [Quit: Leaving]
23:25 -!- louisway [] has quit [Remote host closed the connection]
23:31 * PrinceGuptaGitte sent a long message: < >
23:33 < PrinceGuptaGitte> I can't seem to find a way to map what these values are signifying, like this value is input, this is ouput size etc.
23:34 < PrinceGuptaGitte> Any suggestion is highly appreciated. Thanks
--- Log closed Wed Mar 11 00:00:23 2020