Is RNN double backwards implemented?

Is this implemented? This is a must for papers such as that use an RNN as the critic with an Improved WGAN objective.

If this is not implemented, are there plans to implement this?

If it is implemented it is probably on master only, not on the current version (v0.1.12)