Dual source transformer

I was trying to migrate my model from opennmt-tf to opennmt-py. My tf model use a tf-specific feature (ParallelInputter)
anyone know easy way to implenment this in opennmt-py?

BTW, my 2 sources have same length.

In that case you could merge the 2 sources in a single file and separate words from each source with the special character, e.g.:

she│C was│L eating│L biscuits│L afterwards│L .│N