Adapting Pytorch tutorial “NMT from Scratch…” for dynamic RNN

I want to adapt the tutorial for a dynamic RNN. I have written about the issue extensively on stackexchange. However these forums may be more appropriate home for such a question:

Since I am having heaps of trouble to get this done, I would politely like to ask for any help from the community

kind regards,