DP training for text-to-image diffusion models

T2I models are quite popular at the moment. Differentially private training of T2I models is very useful for a number of domains including healthcare where preserving patient privacy is of utmost importance. Most papers on diffusion models these days use diffusers library from huggingface for implementation. Is there any tutorial on resource that shows how to properly do DP training using opacus and diffusers?