The only real difference between this version and v3.0 is the addition of the Self-Attention Guidance node on the base image generation. This is especially useful with the MoP (Mixture of Perverts) v2.0 model.