00:04 < rcurtin> zax

00:04 < rcurtin> oops, typed into wrong window :)

00:27 < zoq> sumedhghais: What do you mean with strategy?

00:28 < rcurtin> sumedhghaisas: I'll take a look at the GRU tests soon, I needed to think about a response there

00:30 < rcurtin> I think I know what is happening with the FastMKS test---

00:30 < rcurtin> because we are running the tests in parallel now, it's possible that we could run two serialization tests at once

00:30 < rcurtin> but SerializeObjectAll() always saves to the same file... so the file is probably getting overwritten

00:31 < zoq> ah, right

00:32 < rcurtin> I'm opening an issue now, this would be a fun issue for someone new to solve (if anyone has interest, maybe Rajiv might)

00:32 < zoq> yeah, nice idea

00:35 < rcurtin> https://github.com/mlpack/mlpack/issues/1052

07:54 < lozhnikov> kris1: were you able to write the fifth blog post?

07:57 < kris1> Yes i was trying to write yesterday night. But i am confused as this was bad week. I don’t really have much updates.

07:58 < kris1> Still i will push what i wrote right now.

08:41 < lozhnikov> kris1: the header of the file is invalid. The post is not displayed.

08:41 < lozhnikov> The header should look like

08:41 < lozhnikov> Title: <some title>

08:41 < lozhnikov> Date: <some date>

08:44 < lozhnikov> kris1: I have fixed the header

16:07 < partobs-mdp> zoq: I'm almost done with fixing issues from your review. However, I want to ask a question: should we use the separator at all? So far I can't see a use case where it would give any profit. What do I miss?

16:11 < zoq> partobs-mdp: In my experiments, the separator made a difference, so I'd like to keep the option to use one.

16:14 < partobs-mdp> zoq: I see, but what about the SortTask? Does it make difference there?

16:15 < zoq> Or do you think the effort to normalise the input is way more difficult to do with a marker?

16:16 < zoq> I haven't tested the sort task yet, I can run some experiments tomorrow.

16:21 < zoq> Also I think you should update your branch and use the latest rnn class, that could improve the results, since it uses a new initialization method.

16:27 < partobs-mdp> zoq: Almost made it, but rnn.hpp is in my way:

16:28 < partobs-mdp> CONFLICT (content): Merge conflict in src/mlpack/methods/ann/rnn_impl.hpp

16:28 < partobs-mdp> error: Failed to merge in the changes. \n Patch failed at 0022 Fixed LSTM baseline bug

16:28 < partobs-mdp> How to resolve the conflict in that file?

16:31 < zoq> I see, in case the auto merge failed you have to manually fix the issue and continue the rebase. I think an easy solution would be to copy the latest rnn class and just merge the change you did back in.

18:07 < kris1> lozhnikov: just saw your message…… thanks…….i will take care in the future

18:07 < kris1> for the header of the blogpost