What is the function of lightning lora during training? #3
Description
Thanks for your great work!
You said:
Please note that our style transfer model is based on Qwen-Image-Edit-2509, and has to be used with Qwen-Image-Lightning Lora, which we have converted to Diffsynth format for compatibility. Otherwise, the model may suffer from either low-speed or low-quality.
May I ask what is the function of Qwen-Image-Lightning Lora? As far as I know, Qwen-Image-Lightning Lora is merely used for acceleration. Does it can increase the quality of image in your work, or just merely to accelerate?
How do you use Qwen-Image-Lightning Lora during training, or just use it during inference?
Looking forward to your answer.
Thanks a lot!