I suspect it’s not that similar. The intuition behind LoRA is more true the higher the rank of the weights of the model. Even the smallest LLMs have considerably higher rank weights than Stable Diffusion. They are large, after all.