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/
< himanshu_pathak[>
I think gitter bridge is broken we are not receiving messages of this channel on gitter
< HimanshuPathakGi>
Hey @saksham189 @zoq was suggesting for kmeans so I will try to find apply this today. I don't know why we are not receiving messages from riot I think gitter bridge is broken
< nishantkr18[m]>
zoq: Thanks. everything works now.. shall I add the notebook to mlpack/examples ?
< nishantkr18[m]>
zoq: BTW, I was thinking of adding it a separate folder named 'reinforcement_learning', as we'd be adding other notebooks there
< nishantkr18[m]>
zoq: Also, can u point out where http://lab.mlpack.org/ is mentioned? If not, shall we add it to the README.md maybe? So not newcomers can know that they can run the code online.
< HimanshuPathakGi>
kartikdutt18 (Gitter): I don't have an issue with it; we used the local processing idea because shrit pointed out that using other servers can help save any bandwidth costs
< HimanshuPathakGi>
@kartikdutt18 Message for you from rcutin
< kartikdutt18Gitt>
Ahh, Let me use FreeNode. Thanks @himanshupathak21061998.
kartikdutt18 has joined #mlpack
< kartikdutt18>
Hey @rcurtin, Sorry I couldn't reply yesterday, the messages got lost.Let me know what you think about this, @zoq suggested to use boost::iostream::filter to unzip files. That does work but only for zipped files that contain a single file. So inorder to use that we would need seperate zip files for train and tests for datasets like MNIST and Pascal
< kartikdutt18>
VOC as they have fixed training and testing set. An alternative would be adding a dependency such as libarchive. I think it might be better to save as much bandwidth as we can because each run on Azure pipelines leads to a downloads from mlpack.org. As one day, we might hopefully get many PRs on models repo and each commit will lead to a download
< kartikdutt18>
of either datasets or weights.I plan on adding the Pascal VOC dataset, weights for LeNet, darknet and YOLO models. Let me know what you think and I can accomodate the changes for that.Thanks a lot.
ImQ009 has joined #mlpack
togo has joined #mlpack
favre49 has joined #mlpack
< favre49>
nishantkr18[m]: Yup go ahead and make a PR
< favre49>
Perhaps it makes sense to put it under a Q learning folder instead of RL? Looking at how its structured now
favre49 has quit [Remote host closed the connection]
< zoq>
nishantkr18[m]: Great, I agree with favre49, to use q-learning in the name.
< zoq>
nishantkr18[m]: Also yes let's update the examples README, there is a website update in the pipline where this is going to be mentioned as well.
kartikdutt18 has quit [Ping timeout: 245 seconds]
< nishantkr18[m]>
zoq: favre49 Done 👍️
kartikdutt18 has joined #mlpack
kartikdutt18 has quit [Ping timeout: 245 seconds]
< chopper_inbound[>
How can we skip ci testing? I tried appending [ci skip] in commit message but it didn't worked. Am I doing something wrong or ci skip feature is not enabled?
< zoq>
chopper_inbound[: That should work, but not for the jenkins jobs.
< zoq>
chopper_inbound[: done, will go through the code later
< chopper_inbound[>
zoq: Thanks :)
kartikdutt18 has quit [Remote host closed the connection]
kartikdutt has joined #mlpack
kartikdutt has quit [Ping timeout: 245 seconds]
ImQ009 has quit [Quit: Leaving]
< rcurtin>
bad news everyone! many of our build systems have finally gone down (almost two years after I left the company that was hosting them and they forgot about them)
< rcurtin>
this means that there will have to be a good bit of work to set things back up, and we'll have far fewer resources to build with; I think we'll be much more dependent on Azure pipelines as a result
< rcurtin>
goodbye masterblaster; you served us well :)
< zoq>
rcurtin: :(
< rcurtin>
we lose those cool sparc64 servers too, too bad
< zoq>
noooo
< rcurtin>
only have the single sparc64 desktop under my desk now... last I compiled mlpack on it I think it took a full day :)
< zoq>
benchmark nodes as well?
< jeffin143[m]>
So apart from azure ?? The other build were running on masterblaster ???
< rcurtin>
yeah, the benchmark nodes are down too
< rcurtin>
unfortunately there's no point of contact that I have inside of symantec anymore, so there's no chance of getting them back up; I don't think anyone in the company even knew what the systems were doing after I left, just forgotten in a server room somewhere
< rcurtin>
there is one system that should come back up, which is hosted at Georgia Tech, but it's not very powerful
< rcurtin>
there will still be some build capacity of course, it just won't be as insane as the old setup was
< zoq>
I have two machines I can add to the pool, but yeah not as powerful.
< zoq>
i7-7700 + 64GB memory, I think masterblaster had 256GB memory
< rcurtin>
yeah, first I think I need to set some things up differently, like the docker registry on mlpack.org instead
< rcurtin>
give me a day or two to think about what the right thing to do here is
< rcurtin>
(it's also possible the systems will magically come back up! but it seems unlikely)
< rcurtin>
for now I'll at least get a copy of armadillo-8.400.0.tar.gz hosted elsewhere
< zoq>
I guess good thing shrit is work on getting mlpack to run on constrained devices :)
< zoq>
rcurtin: sounds good
< rcurtin>
:)
< rcurtin>
I'm disabling all the matrix builds for now too, so that things don't get overloaded
< jjb[m]>
rcurtain: I think now would be an optimal time to switch to GitHub Actions 🙂
< rcurtin>
jjb[m]: yeah, I don't disagree; I don't know how that integrates with Jenkins
< rcurtin>
for now I don't think I have time to investigate too much though
< zoq>
jjb[m]: If I remember right github actions builds on azure pipelines.
< rcurtin>
and honestly, when I reflect on it, a lot of the builds we were doing were a bit overkill---we tested many different configurations every night
< rcurtin>
that was useful for finding flaky tests, but maybe we can come up with a more efficient way
< zoq>
agreed
< jeffin143[m]>
Zoq , rcurtin : if you declare arma::mat<unsigned char> matrix
< jeffin143[m]>
And store values
< jeffin143[m]>
And the print cout<< matrix
< jeffin143[m]>
It prints ok
< jeffin143[m]>
But if your of for I =0 and print it prints absurd char
< jeffin143[m]>
Any idea why ???
jeffin143 has joined #mlpack
< jeffin143>
arma::Mat<unsigned char> matrix; data::ImageInfo info; data::Load("test_image.png", matrix, info, false) ; cout<<matrix<<endl; for (size_t i = 0; i < matrix.n_elem; ++i) { cout<<matrix[i]<<","; }
< rcurtin>
I don't understand "But if your of for I = 0", can you explain?
< jeffin143>
They have handled like this in python
< jeffin143>
I am not sure how shoudl we handle it in C++
< rcurtin>
probably std::cout is not the right way; that's for formatted output
< rcurtin>
I think that you want unformatted output
< rcurtin>
i.e. "binary" output
< jeffin143>
I was printing it to store it in array , I mean I will write it to a file and copy paste that array ,
< jeffin143>
Yes I want unformatted output
< jeffin143>
How should I get that
< rcurtin>
I would suggest std::fstream or std::ostream
aoeuiiueoa[m] has joined #mlpack
< jeffin143>
Ok I thought cout worked exactly same as filestream and hence I tried with cout and was check if it worked in console most probably it will work when I write to file
< jeffin143>
I ll use and fstream and return back
< shrit[m]>
rcurtin Today I told myself, Why I am not able to download armadillo and boost from masterblaster? maybe it is down, or I do not have the right to access. So I downloaded the depdencies from their website, bad news to hear that we lost masterblaster
< rcurtin>
yeah, it's unfortunate, but I knew it was coming sooner or later :)
< rcurtin>
I feel happy we were able to use it for as long as we could
< shrit[m]>
Actually, we might need to host armadillo somewhere else other than souceforge, the download rates for the server are very slow
< rcurtin>
yeah, I just put 8.400.0 on ratml.org for now (same server as mlpack.org)
< shrit[m]>
I was thinking about older version less than 5, Are they still used today, or supported?
< rcurtin>
no, Conrad pushes pretty hard to get people on new versions
< rcurtin>
the only reason our minimum is 8.400.0 is to support people on old Ubuntu who want to just do `apt-get install libarmadillo-dev`
< rcurtin>
(or old RHEL)
< rcurtin>
if there's still code in the codebase for older versions, please feel free to strip it out :)