Does RDM training use x0-parameterization?

I am training rdm and I noticed the code uses x0-parameterization, is this the same as the rdm training in paper?

Does eps-parameterization harms performance comparing to x0-parameterization in rdm training?

Thanks!