This is probably related to this question: How to make a seq2seq model work with infinite vocabulary?
I have read many seq2seq implementations and they all seem to only work on fixed, well-discretized vocabulary which the main application of seq2seq have them.
I'm trying to use a seq2seq model for things similar to a style transfer task which involves temporal sequences / real number features. Do I have to discretize the time into log-scale windows accepting information loss, or there are brilliant ways to make infinite vocabulary working with seq2seq?
