11:47 < partobs-mdp> zoq: back home, now implementing the AddTask/SortTask change you requested (moving reshape inside task defiinition)

11:48 < partobs-mdp> found two more nasty bugs:

11:48 < partobs-mdp> - mlpack_baseline was running on 20 epochs no matter what the value of -i parameter is

11:49 < partobs-mdp> - SortTask receives binary representations without any delimiters - how on earth the network is supposed to know where one number ends and another begins?

11:49 < partobs-mdp> Sure enough, I fixed first bug - the second bug is more tricky, tho

11:52 < zoq> partobs-mdp: Great, I'll take a look once you pushed the changes. About the second issue, I've included an end marker for the copy repeat task, see my last comment on the PR.

11:54 < partobs-mdp> zoq: Pushed, I'll look at your PR.

12:10 < partobs-mdp> zoq: Read your comment. I don't quite understand what kind of marker did you use. In AddTask it was a special symbol [0 0 1] which didn't correspond to neither 0 nor 1. What about CopyTask? (and thus, SortTask)

12:12 < partobs-mdp> By the was, is it okay to write off-topic messages here?

12:12 < partobs-mdp> (I mean, about something loosely related with machine learning research, but not with mlpack)

12:13 < zoq> partobs-mdp: Of course!

12:14 < zoq> partobs-mdp: I used 0.3 as end marker, set the marker after the normalization step.

12:15 < zoq> I haven't tested it without the marker.

12:15 < partobs-mdp> Okay then :) While away from home, I have written a small tool to fetch all arxiv articles from email digest to my home dir (abstract.txt + paper.pdf). Here it is: https://github.com/partobs-mdp/arxiv-fetcher If anyone is interested, test it, use it and spread the word :)

12:16 < partobs-mdp> (Disclaimer: I didn't search for tools of this kind - rather, I wanted to make something for <1h that would be useful)

12:17 < partobs-mdp> (Disclaimer 2: I use Ubuntu, thus the script is not entirely portable - I need some OS-agnostic notifier system to rectify this)

12:18 < zoq> partobs-mdp: Nice! I'll test it out.

12:18 < partobs-mdp> zoq: You mean, like this: [[1,0],[1,0],[0.3, 0.3],[0,1],[0,1]]?

12:19 < partobs-mdp> zoq: About that script: even though I pose it as a personal script, feel free to open an issue if something goes wrong or not optimal ;)

12:19 < zoq> I used the end marker only for the first input, but yes.

12:20 < partobs-mdp> zoq: You mean that you marked the *beginning* of the sequence iteration?

12:20 < partobs-mdp> [[1,0.3], [1,0], [0, 1], [0,1]]?

12:20 < partobs-mdp> Or did you put it in the response sequence?

12:20 < zoq> I mean I used: [[1,0],[1,0],[0.3, 0],[0,1],[0,1]]

12:21 < zoq> the first is the input and the second is the repeat information right?

12:21 < partobs-mdp> I see :) How did you do for repeats? Like this: [[1,0],[1,0],[0.3, 0],[0,1],[0,1], [0.3, 0], [0,1], [0,1]]?

12:21 < partobs-mdp> zoq: yes, that's how we arrived at that representation

12:23 < partobs-mdp> zoq: by the way, was there any special reason for using 0.3, and not (say) 0.5?

12:25 < zoq> I just set the end marker once, so for repeat = 2; so if my sequence I like to copy looks like 0 0 1 I just do 0 0 1 0.3 and the repeat information looks like 0 0 0 0 1 1 1 1 1 1.

12:26 < zoq> Not not really, I guess you could test another value for the end marker the tricky part is to choose something that is kinda unique after the norm step.

12:27 < zoq> I guess, we could norm first and choose the marker afterwards, that might be clever.

12:29 < zoq> Btw. usually I use arxiv-sanity, but I always wanted something that downloads the files for me.

19:44 < zoq> rcurtin: Somehow missed the dicussion here: https://github.com/mlpack/mlpack/pull/1038#issuecomment-311096906 I'll take a look at it tomorrow.

19:45 < rcurtin> I know the feeling, I seem to miss a lot more discussions these days :)

22:38 < chenzhe> rcurtin: Are you there?

22:40 < chenzhe> Hi Ryan, you mentioned Jenkins in your comments, can you explain to me a little bit how to use it?

23:42 < rcurtin> chenzhe: hi there, let me reread my comment so I can remember what I said ...

23:43 < rcurtin> ah, right, so the style checker runs automatically when you push code to the PR

23:43 < rcurtin> you can see at the bottom of the PR it says "N out of 3 checks successful" (or something similar)

23:43 < rcurtin> so the three checks are the style check, a linux build test, and a windows build test

23:44 < rcurtin> you can click "details" on the "Style Checks" to see more information

23:44 < rcurtin> in the case of your PR #1041, the URL is http://masterblaster.mlpack.org/job/pull-requests%20mlpack%20style%20checks/400/

23:44 < rcurtin> and if you scroll to the bottom of that page you can see "Cppcheck results", and you can click on those to see what the issue is

23:45 < rcurtin> in your case it looks like the issue is very simple, "missing space after comma" :)