Stable Diffusion Masterclass - Theory, Code & Application

HuggingFace Diffusers Library - Inference with Trained LoRA Model

Now that we've seen how easy it is to use a trained textual inversion embedding with a Diffusers pipeline, let's see how we can use a trained LoRA model with a Diffusers pipeline.

We earlier learned that, like textual inversion, LoRA is another method that we can use to introduce ourselves, new concepts, and styles to Stable Diffusion. Specifically, with LoRA, we insert and train new layers on top of the base diffusion model to learn the new concept or style. We can then use the model along with its new trained layers to generate images of the new subject.