How to Load LoRA Weights into UNet2DConditionModel Individually?

[ancientluna]

I’ve created a custom pipeline and am not using StableDiffusionPipeline. I’m wondering if there’s a function similar to load_lora_weights() that would allow me to load LoRA weights in .safetensors format into the UNet2DConditionModel individually? Any guidance would be appreciated!

[ancientluna]

I think I succeeded in using unet.load_lora_adapter(). Anyway, since my LoRA weight was saved by stablediffusionpipeline, I had to modify the key when loading it with unet. I'm wondering if this is the best approach?