Diffusion text-to-text models

Hi everyone :wave:

I noticed that there is a lot going on the last year in the NLP/diffusion models space like DiffuSeq, TESS, or Diffusion-LM. Is there someone in the Diffusers Community already experimenting with these models and can give me a hint to get started? Are there even plans to incorporate text-to-text models into Diffusers? I find the topic and the advantages of diffusion models with respect to text rather fascinating and would love to dive deeper, but I do not have the time and resources to start from completely from scratch by reproducing the stuff from the original papers/authors. Any response/input would be highly appreciated.


1 Like