Nice work! I have a question, in this paper, cfg is decoupled with the distilled model, and therefore during inference, the distilled model will get 2 NFE in each step. "On Distillation of Guided Diffusion Models"(https://arxiv.org/abs/2210.03142) provides a method to do the CFG distillation to remove the extra NFE, have you ever compare with the method? Thanks!