Seq2Seq tasks with Diffusion LM

[chiral-carbon]

Hi,

I was wondering how I might go about implementing a machine translation task with DiffusionLM. I have trained DiffusionLM for unconditional generation and conditional generation for a few tasks, but I want to be able to use it for translation tasks.
Wanted to ask so that I can get some insight. Would it require training a translation based classifier to guide the model?

Would appreciate any advice, thanks!