provide API for returning output from intermediate layers

[zhouyong64]

It would be very helpful to have an API for returning output from intermediate layers, for example, the one before the final layers. This output can be used in other speech tasks other than speech recognition.

[daanzu]

Interesting idea. This would be a bit complicated, because I think we would need to build the torchscript model differently to allow for this, in addition to modifications to the library code. My plan was to keep this library fairly simple, but I am not entirely opposed to adding this. I may not get around to adding it myself, though. Leaving open.