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/
travis-ci has joined #mlpack
< travis-ci> mlpack/ensmallen#1137 (restore-original-license - d138480 : Ryan Birmingham): The build passed.
travis-ci has left #mlpack []
Prometheus[m] has joined #mlpack
< Prometheus[m]> Hi, I am Karthikeyan. I was looking the ideas page for Gsoc 2021 and noticed visualization tool as an idea. If I am correct isn't there already mlboard for that?
< zoq[m]1> Prometheus there is but there is still some things to do, to use everything Tensorboard provides.
< Prometheus[m]> zoq Is it still available or is someone working on it?
< Prometheus[m]> Gopi Manohar Tatiraju Thank you
ImQ009 has joined #mlpack
TathagataRahaGit has joined #mlpack
< TathagataRahaGit> Hi everyone!
< TathagataRahaGit> I am Tathagata, a third-year UG student at IIIT Hyderabad and I was interested in contributing to mlpack for GSOC. I was thinking of a project related to NLP as I have been working on that field for the last year.
< TathagataRahaGit> Last year, I saw that there was a project on implementing BERT in mlpack
< zoq> TathagataRahaGit: Hello there, BERT is still open if you are interested.
< TathagataRahaGit> (edited) ... in mlpack => ... in mlpack. I was thinking of continuing it and build other transformers models like xlnet and roberta. Also we can have something like the pre-trained models like HuggingFace has.
< zoq> Sounds good, having pre-trained models makes sense to me especially in the NLP area.
< TathagataRahaGit> Another interesting thing that I noticed(although I am not quite sure), is LSTM has not been implemented. I was looking at the examples, specifically https://github.com/mlpack/examples/blob/master/lstm_stock_prediction/lstm_stock_prediction.cpp and https://github.com/mlpack/examples/blob/master/lstm_stock_prediction/lstm_stock_prediction.cpp in which RNNs have used in the name of LSTM
< TathagataRahaGit> So, it would also be a nice idea to implement the LSTM and GRU layers which are very important to NLP
< TathagataRahaGit> Sorry, I have missed it there then.
< TathagataRahaGit> Also, I am new here. Is it too late to start contributing?
< zoq> No worries, not at all Google still has to announce the orgs for this year.
< TathagataRahaGit> Also, I tried joining the slack from the community tab but the link https://mlpack.slack.org looks like it is invalid.
< rcurtin[m]1> oh, right, it's slack.com... let me fix that now, thank you for pointing it out!\
< rcurtin[m]1> ok, pushed a fix; the page should be updated in a few minutes 👍️
< TathagataRahaGit> Also, in order to create an account, I was trying to access this http://slack-inviter.mlpack.com/ but this also looks unreachable
< rcurtin[m]1> yeah, it's on port 4000: http://slack-inviter.mlpack.org:4000/
< rcurtin[m]1> (the URL if you click it actually links to that, but let me also be explicit about the port in the listed link)
< rcurtin[m]1> by the way, little fixes like that are hugely important, so if you see more, do feel free to point them out :) they're very easy to overlook
< rcurtin[m]1> 👍️
< jonpsy[m]> Hey there, I'm NG Sai from IIIT SriCity. I've worked extensively on shogun repository and had a lot of fun. I'm planning to participate in GSoC in this repo, as I'm the repository seems to be silent for quite a while. Hi all!
< jonpsy[m]> * Hey there, I'm NG Sai from IIIT SriCity. I've worked extensively on shogun repository and had a lot of fun. I'm planning to participate in GSoC in this repo, as the repository seems to be silent for quite a while. Hi all!
< jonpsy[m]> * Hey there, I'm NG Sai from IIIT SriCity. I've worked extensively on shogun repository and had a lot of fun. I'm planning to participate in GSoC in this repo, as that repository seems to be silent for quite a while. Hi all!
< rcurtin[m]1> Hey jonpsy , I lurk in the Shogun channels and have seen your name there :) I think you were contributing there for quite a long time, if I remember right 😃 welcome!
< jonpsy[m]> Yeah, I'm glad you noticed :). Its a shame that its inactive, I was really all-set for GSoC this year with that repo.
< jonpsy[m]> Mlpack is a cousin repository, so it shouldn't be too difficult :)
< rcurtin[m]1> Heiko and Viktor are still out there, but I think busy with their jobs these days
< jonpsy[m]> Indeed, but i haven't seen them posting any updates for GSoC. Regardless, better safe than sorry.
< jonpsy[m]> Besides, mlpack is a cool repo :D
< rcurtin[m]1> 👍️
< shrit[m]> @rcurtin I have some more details about the issue
< shrit[m]> I will paste the out put here of the simple train test
< shrit[m]> armadillo version: 9.800.3 (Horizon Scraper)
< shrit[m]> Create Children from91points, near set size13far set size65used set size1
< shrit[m]> Create Children from50points, near set size1far set size7used set size1
< shrit[m]> Create Children from76points, near set size7far set size44used set size1
< shrit[m]> Create Children from32points, near set size1far set size1used set size1
< shrit[m]> Create Children from27points, near set size1far set size4used set size1
< shrit[m]> Create Children from30points, near set size1far set size11used set size1
< shrit[m]> Create Children from17points, near set size5far set size10used set size1
< shrit[m]> Create Children from4points, near set size2far set size0used set size1
< shrit[m]> Create Children from6points, near set size1far set size1used set size1
< shrit[m]> Oops sorry, It did not paste all the output
< shrit[m]> So when it arrives at the last one (4 points). The RPI stops outputting, and the memory start building up megabytes until it is saturated after 5 seconds
< shrit[m]> I have the same output at each run
< rcurtin[m]1> oh my! any chance you can get a debugger on it and ctrl-c when the output stops to get an idea of what's going on?
< rcurtin[m]1> given that output, it looks like the tree building procedure itself might be terminating fine
< shrit[m]> Yes of course, I have a debugger
< zoq> shrit[m]: I'll setup the key in the next minutes.
< shrit[m]> Thanks
TathagataRahaGit has left #mlpack []
< zoq> shrit[m]: Just send you the details for viking.
< shrit[m]> Thanks, is there any gdb installed?
< zoq> Now there is.
< shrit[m]> I got the same output from GDB, even viking is blocking when arriving at 4 points
< shrit[m]> Is there any dubugging sybomols or do I leave it until it stops automatically? I have to use Ctrl_C therefore no stack can be traced
< shrit[m]> * Is there any debugging symbols or do I leave it until it stops automatically? I have to use Ctrl_C therefore no stack can be traced
< rcurtin[m]1> you'd need to compile with `-DDEBUG`, then ctrl+c will show the right backtrace
< shrit[m]> 👍️
< rcurtin[m]1> just added another arm64 build node to Jenkins: http://ci.mlpack.org/computer/pop/
< rcurtin[m]1> it uses the label `linux-arm64`, and I updated the other two arm64 nodes to have that label also
< rcurtin[m]1> shrit: does your RPi have a hostname? we can change the name from `arm64` 😃
< shrit[m]> unfortunately my RPI hostname is ubunut, which is not helpful, and the domain name is at my name. You can call it any name you want, or follow an existing convention
< shrit[m]> * unfortunately my RPI hostname is ubuntu, which is not helpful, and the domain name is at my name. You can call it any name you want, or follow an existing convention
< shrit[m]> Maybe use periodic table element for machines;
< rcurtin[m]1> it's your choice 😃 your system so you get to choose whatever name you like 😃
< shrit[m]> poor machine, it is consuming 32GB of ram + 16GB of swap.
< shrit[m]> It does make difference using DDEBUG
< rcurtin[m]1> ouch 😃
< shrit[m]> @zoq, I am not able to ssh to vikings, password is not matching public key
< zoq> Hm, okay, let me check.
< shrit[m]> i was able, and then it stopped
< zoq> shrit[m]: can you try again?
< shrit[m]> It is working now thanks
< zoq> Let me know if it fails again.
< shrit[m]> I am waiting on transferring the mpack_test to the viking ETA 2 hours
< zoq> Are you uploading from your local machine?
< shrit[m]> I do not think that viking is able to compile mlpack,
< shrit[m]> yeap
< zoq> I build mlpack wihtout debug symbols just fine.
< zoq> I guess, if you build with -j4 it will be faster than uploading.
< shrit[m]> Yes, but It consumed 32GB + 16 GB for debugging symbols
< shrit[m]> ram + swap
< zoq> with -j1?
< shrit[m]> I will give it a try
< zoq> Yeah, maybe not go for all cores
< zoq> I guess 3 or 2 should be fine.
< shrit[m]> I need to use DDEBUG, so hope memory will stay calm
< shrit[m]> My network is from stone age
< shrit[m]> It is interesting but the aarch64 processor of the viking is still much slower than Xeon at compilation
< shrit[m]> * It is interesting that the aarch64 processor of the viking is still much slower than Xeon at compilation
ImQ009 has quit [Quit: Leaving]
< rcurtin[m]1> shrit: I see you assigned #2827 to yourself (adaboost test failure on aarch64); I have the issue reproduced in a container though (I didn't see the assignment until now)
< rcurtin[m]1> should I keep digging on that issue, or do you want to handle it?