13:28 < zoq> kris1_: Hello, I'm curious what is the status of the RBM PR? It looks like is almost ready?

13:30 < kris1_> No actually i talked with mikhail but we are refactoring both rbm. So that we can push both ssRBM and RBM togther.

13:30 < kris1_> I am almost finished with ssRBM implementation there are some issues i am fixing. And i still have to add the test sorry for the delay

13:30 < zoq> kris1_: Ah I see, so I guess some more days.

13:31 < kris1_> I think i would close the existing PR on RBM since all the changes will be now made to ssRBM PR.

13:31 < zoq> kris1_: No problem I was just looking at the PR, and I thought I ask what the status is :)

15:45 < partobs-mdp> zoq: Could you send me the outputs of the shell scripts from mlpack-models fork we talked about yesterday?

15:48 < zoq> partobs-mdp: I'm on it, once I have the result I get back to you.

15:49 < zoq> partobs-mdp: I was testing the fixed output but run into several issue, like the score function doesn't work with arma::mat.

15:51 < zoq> partobs-mdp: Also sometime back I mentioned that it might be a good idea to normalize the sequence, which isn't easy to do outside the Task, if we use the separator.

15:57 < zoq> partobs-mdp: I think we have to do some additional preprocessing to get some good results with the LSTM model, but let me run the script.

16:23 < zoq> partobs-mdp: First output: https://gist.github.com/zoq/e71171c3ca38c24bfa700e2acc24586b

17:11 < zoq> partobs-mdp:: Second output: https://gist.github.com/zoq/1625cf338d6dca769afd62dbe595ca41

17:33 < zoq> rcurtin: The PR job doesn't use the libs from the checkout job, so I have updated the PR job.

17:50 < rcurtin> zoq: ah, right! thanks for the fix. I am still running the checkout job with 6.0.0 anyway though, so once the tests work it will be ready :)

19:38 < zoq> rcurtin: I think, we can stop the checkout job, with the shogun issue it will run forever ...

19:57 < rcurtin> zoq: right, but at least shogun 6 is installed now :)

19:58 < rcurtin> I guess it is all a bit pointless since I forgot the PR job does not use the checkout though

19:58 < rcurtin> and we'll have to build it again to run the tests anyway once the PR is merged

21:20 < zoq> partobs-mdp: The output of the last script: https://gist.github.com/zoq/7b4fba0869ad4a05cf491513fe518837

22:03 < sumedhghaisas> zoq: Hey Marcus... had some questions about BPTT

23:59 < sumedhghaisas> rcurtin: Hey Ryan... I was just implementing the changes suggested by you. I came accross one really minor problem with that thing. But its too complicated to explain GitHub. It has to do with how BPTT is implemented. Are you free right now?