recurrent dropout

[sencilla]

Hi. I think the recurrent cells in flux need a dropout for their internal states to improve training in larger networks

For example : mycell= GRU(10,10,0.1)

Where 0.1 is dropout

Or is there an easy way to implement this manually?

[CarloLucibello]

Dropout in between stacked layers can be implemented as described in
#2514 (comment)
We should add this to the guide on recurrent layers.