Inference with LoRA weights

[trqminh]

Hi @pixeli99 ,

Thank you for your great work! I am wondering if you have the script for inference with lora model. I am struggling with loading lora and run the trained lora models. It seems I cannot get the pipeline to run the lora part in inference, it just like I am running everything with the original pretrained model.

Thank you for your help!

[zzhang2816]

Add these lines to inference code

if lora_folder != "":
   unet.load_attn_procs(lora_folder)
   unet.to(torch.float16)
   unet.requires_grad_(False)