04:26 < ShikharJ> rcurtin: Can we merge the WGAN PR now? The tests pass, and the review comments have been accounted for.

10:56 < zoq> ShikharJ: What was the issue, not sure I see the solution since the last commit was stashed?

10:56 < zoq> wenhao: Have fun!

11:00 < ShikharJ> zoq: Actually, when we were generating the images in csv, we were just doing generator.Forward(), which worked well, as we were saving just a reference. But after changing to move semantics, this would be wrong, and we'd have to do gan.generator.Forward().

11:01 < zoq> ShikharJ: Ahh, I see, nice catch.

11:06 < zoq> ShikharJ: About parallelization, another idea we could test without any code modification is to use Downpour SGD (https://github.com/mlpack/mlpack/pull/1117), we can run the optimizer on the same system.

11:06 < zoq> ShikharJ: Another idea is to see if we could parallelization some steps using OpenMP. But let us finish the RBM PR first.

11:09 < ShikharJ> zoq: Hmm, agreed. I've updated the changes on WGAN in the meantime. RBM PR currently works for the Stochastic method, so probably next step would be to introduce the tests as Mikhail had mentioned, and then add the support for mini-batches.

11:09 < zoq> ShikharJ: Once the build finishes I'll hit the merge button.

11:10 < zoq> ShikharJ: Agreed, sounds good.

13:03 < rcurtin> ShikharJ: sorry, I didn't see that any of the code had changed, but the serialization looks correct to me now

13:03 < rcurtin> thanks for handling that :)

13:05 < ShikharJ> rcurtin: No worries, I'm glad we were able to complete all of this so soon. All credit goes to zoq, kris, and lozhnikov !

14:19 < rcurtin> don't forget yourself in the credit list, you did a lot of work also :)

14:24 < manish7294> rcurtin: Is it okay to have a quick talk now over the things?

16:22 < zoq> ShikharJ: I agree with Ryan :)

18:16 < rcurtin> manish7294: sorry, I am out all over the place today, could tomorrow work?