It seems like you can apply LCM LoRAs to any fine-tuned version of Stable Diffusion or SDXL without having to distil them and speed up inference.

LoRA Weights: https://huggingface.co/collections/latent-consistency/latent-consistency-models-loras-654cdd24e111e16f0865fba6

Technical Report: https://github.com/luosiallen/latent-consistency-model/blob/main/LCM-LoRA%20Technical%20Report/LCM-LoRA-Technical-Report.pdf

Blog: https://huggingface.co/blog/lcm_lora