09:28 < partobs-mdp> zoq: rcurtin: Wrapped up all unit tests for benchmarking (in the corresponding PR). However, there is a merge conflict. How can I resolve it?

09:32 < zoq> partobs-mdp: Rebase the PR against the current master branch, or you can also click on "Resolve conflicts" on the PR and manually fix the issue. I can also fix the issue for you, just let me know.

09:49 < partobs-mdp> zoq: Rebased, now waiting for running checks (Jenkins still crashes)

10:19 < ironstark> rcurtin: zoq: I am facing issues while writing dlib-ml scripts

10:19 < ironstark> this is my KMEANS script

10:20 < ironstark> can you please tell me the issues. ( I know there are many :( ).

10:20 < ironstark> https://www.irccloud.com/pastebin/ZNAys3Bc/KMEANS.cpp

10:21 < ironstark> https://www.irccloud.com/pastebin/0XFouNBN/KMEANS.py

10:22 < ironstark> This is the example file I took help from http://dlib.net/kkmeans_ex.cpp.html

10:29 < zoq> partobs-mdp: About the jenkins issue, Saurabh and Ryan are working on the jenkins system, so they might be ale to provide some helpful input, so let us wait some more time, before I go and take a closer look.

10:30 < zoq> ironstark: Let me take a look.

10:40 < zoq> ironstark: There is no direct conversion between arma and dlib, so you can't just write: std::vector<sample_type> samples = referenceData;

10:41 < zoq> ironstark: I sketched something up here: https://gist.github.com/zoq/c8e2575d56344c3924cc1d547bb77ecd

10:41 < zoq> ironstark: I also I changed the dimension for: typedef matrix<double,referenceData.n_rows,referenceData.n_cols> sample_type; the second dimension should be 1, since referenceData.n_cols represents the number of samples so referenceData.n_cols = samples.size()

10:41 < zoq> ironstark: I couldn't find a centroids file/matrix, maybe I missed something?

10:42 < zoq> ironstark: We have to do the same conversion as we did for the samples.

10:46 < ironstark> zoq: Yes I haven't built the centroid matrix yet. Will do the same. Thanks a lot for the help.

10:47 < zoq> ironstark: Ah okay, let me know if you run into any other issues, I haven't tested the code, so there might be some issues that I missed.

10:48 < ironstark> sure

13:38 < lozhnikov> kris1: hi, were you able to try the ssRBM on the CIFAR dataset?

13:39 < lozhnikov> kris1: and were you able to write the 9th blog post?

13:39 < kris1> Hi, Sorry i wasn’t able to work on this weekend as busy with some administrative works in my institute.

13:39 < kris1> I would working on it today.

13:40 < lozhnikov> okay, no problem. And what about the blog post?

13:40 < kris1> I was able to complete gan with only one optimizer though. Yes i would write the blogpost tonight. Or tommrow in the morning.

13:40 < lozhnikov> ok

13:41 < kris1> The gan is giving some assertion errors that i was trying to figure though due to lack of time i was not able to do so. Maybe later today i can update the PR.

13:42 < lozhnikov> sounds good

13:47 < partobs-mdp> zoq: Fixed everything from your review. When are we going to merge the benchmark PR - I'm feeling a bit tight on time :|

14:09 < zoq> partobs-mdp: Okay, let me take a final look.

14:16 < zoq> partobs-mdp: Once we fixed the style errors the PR should be ready, at least from my side.

14:35 < ironstark> zoq: looks like we cannot use referenceData.n_rows here :

14:35 < ironstark> typedef matrix<double, referenceData.n_rows, 1> sample_type;

14:35 < ironstark> any ideas about what can we do?

14:36 < ironstark> I tried placing referenceData.n_rows in a const int variable and then use that variable in matrix

14:36 < ironstark> but that does not work too

14:36 < ironstark> one other option is setting it as typedef matrix<double, 0, 1> sample_type;

14:37 < ironstark> and specify the size at runtime but I am not sure about how to achieve that

14:42 < zoq> ironstark: Ah, right it must be known at compile time. You can set the size at runtime using m.set_size(referenceData.n_rows);

15:04 < ironstark> zoq: other things are working fine now. Just one error in the make file /usr/bin/ld: cannot find -ldlibml

15:07 < ironstark> sorry, not that ..

15:07 < ironstark> https://www.irccloud.com/pastebin/XXLEranw/

15:09 < zoq> ironstark: Can you post the command from the Makefile?

15:11 < zoq> Maybe it helps do link against mlpack or dlib first, so swapping the order, not sure that would solve the issue.

15:12 < zoq> If not, can you open a PR with the existing code?

15:17 < ironstark> https://www.irccloud.com/pastebin/o26zylsC/

15:17 < ironstark> zoq: this is the command from the makefile

15:19 < zoq> ironstark: ah, right, does '-ldlib -lmlpack -lboost_program_options' work?

15:39 < ironstark> zoq: No same error

15:39 < zoq> ironstark: Okay can you open a PR?

15:39 < ironstark> sure

15:40 < zoq> ironstark: Okay great, thanks!