Does the PyTorch version support residual connection?

Dear all,

I am wondering if the PyTorch version supports residual connection on the RNN encoder.
I found some discussion about this going back to June 2017 when it was not supported.

However, I am wondering if it has been added in the meanwhile.