Colin Raffel

I think this should be possible by setting `learn_init=True`, training, and then at test time simply feeding in an input of all zeros, right?

Ah, yeah, I see what you mean. Yeah, I'm not thinking of a simple way to do this with the code as it is, because it treats the input sequence...

In that case, I think this would make most sense as an example script, as I don't think any functionality needs to be added to the LSTM/recurrent layer classes themselves.

> As far was I can see the timestep-by-timestep strategy is not possible with LSTM layers, because there is no way to get the cell state and manually carry it...

> That is not an existing feature, right? Sure it is, `cell` is a public attribute of every `LSTMLayer` instance. > Well, in my personal framework I don't create a...

Oops, sorry, you're right! Currently working on other stuff so not thinking about this straight. Although I'm a little confused why you want to get the cell state instead of...

I think this is a matter of convention, i.e. if a note falls entirely within a sampling interval should it be included or not? I would welcome a PR to...

I'm not sure where I read that all meta events (including lyrics events) should be on track 0. Feel free to make a PR.

I'm not aware of many pieces of music where different singers are singing different words at the same time, but if this ever happens I agree that making it part...

That makes sense. I won't be able to make these changes myself but if someone makes a PR I can review.