07:25 < AyushSingh[m]> zoq thinking of continuing with the implementation of seq2seq sometime in the future(as seemed a little complicated).

07:25 < AyushSingh[m]> For now, I noticed that we only have LSTM and fast_lstm, but not bidirectional versions of these. So, should I look into how to add bi-lstm to mlpack?

07:41 < AakashkaushikGit> Hey @Ayush Singh i am not really sure but can we not import lstm, gru, fast_lstm into the brnn.hpp which i believe is just a container for bidirectional recurrent networks.

09:16 < AyushSingh[m]> Thanks Aakash Kaushik, I didn't notice this earlier.

17:43 < AakashkaushikGit> Hey does anyone have a idea where the `IdentityLayer` layer in the `ann_layer_test` come from ? I am having a hard time finding it

18:17 < zoq> AakashkaushikGit: It's a typedef - https://github.com/mlpack/mlpack/blob/master/src/mlpack/methods/ann/layer/base_layer.hpp#L152

19:04 < AakashkaushikGit> Thanks @zoq :D

19:19 < AakashkaushikGit> Btw @zoq as you mentioned in your latest PR we haven't yet decided to port `base_layer.hpp, refactoring tests is a bit hard in that sense because in the `ann_layer_test` just the identityLayer is used 40 time and so i believe the other typedefs are also used in the tests, so what do we do about it ?

19:20 < AakashkaushikGit> (edited) ... port `base_layer.hpp, refactoring ... => ... port `base_layer.hpp`, refactoring ...

20:23 < jeffin143[m]> @zoq is in beast mode with the migration

20:23 < jeffin143[m]> Do you need some help ?

21:11 < zoq> AakashkaushikGit: I guess we can port the base layer and stick with the design for now.

21:12 < zoq> jeffin143[m]: haha, it's not super fun, similar to the catch refactoring :)

21:13 < zoq> jeffin143[m]: If you like to pick a layer, feel free :)