Rasul Karimov

@pskrunner14 And also on this. Whom should I contact?

@pskrunner14 should I pass an additional argument to the `attention.forward` method or It will be more clear if I create separate classes for different attention models and keep single base...

@pskrunner14 I can work on this if you need some help.

Yeah, I can think of this, will post the solution here.

Also why not renaming `initialize_state` to `initialize_sampling_state` (also add additional `state` parameter) to remove redundant code in `build_logp_and_deterministic_functions` with all the state/observed checks?

Working on it, guess there are some related issues to fix first.

@ColCarroll, yes, the third one is, indeed, looks little tricky with my knowledge of the current design requirements. So yeah, change what you need to.

Sure, I will do it in couple of hours. I'm not sure about the Mixture because it is in progress in different PR. I've just added Mixture for the SMC...

Let me add some tests for sMC and then open a separate PR. `batch_stack` is killing me right now. Especially `.sample()` method which doesn't care about the plate variable.

@junpenglao I kind of forgot about that comment, will test it.