This is actually a training adaptor for z-image that you apply during training to reduce de-distillation. Idea is if you de-still z-image turbo prior to training, then the lora trained on top of that training will de-still much slower.

However it can be used also to de-distill z-image base (which might also be useful for other purposes)