This is an experiment to de-distill guidance from flux.1-dev. We removed the original distilled guidance and make true classifier-free guidance reworks.
Following Algorithm 1 in
On Distillation of Guided Diffusion Models, we attempted to reverse the distillation process by re-matching guidance scale w. we introduce a student model
x(zt) to match the output of the teacher at any time-step t ∈ [0, 1] and any guidance scale w ∈ [1, 4]. We initialize the student model with parameters from the teacher model except for the parameters related to w-embedding.
Since this model uses true CFG instead of distilled CFG, it is not compatible with diffusers pipeline. Please use
inference script or manually add guidance in the iteration loop.
Train: 150K Unsplash images, 1024px square, 6k steps with global batch size 32, frozen teacher model, approx 12 hours due to limited compute.
