view reply I would say it should work for Schnell but you might have to change the loss function a bit to account for the fact it's a timestep-distilled model. Yes. I don't think LoRA merge should increase the final state dict size. The numbers shouldn't change much.
view article Article (LoRA) Fine-Tuning FLUX.1-dev on Consumer Hardware By derekl35 and 4 others • 7 days ago • 62
view reply https://github.com/huggingface/diffusers/tree/main/examples/research_projects/flux_lora_quantization Was provided by SimpleTuner group?
view article Article (LoRA) Fine-Tuning FLUX.1-dev on Consumer Hardware By derekl35 and 4 others • 7 days ago • 62
Fine-Grained Perturbation Guidance via Attention Head Selection Paper • 2506.10978 • Published 13 days ago • 26 • 3