result images is blur with lora

[WM-Wu]

Hello, I use my own data to train to get lora and reason, and then find that it is easy to have fuzzy results, and sometimes there are animation pictures that I do not want. I adjusted parameters such as rank or step_num and it didn't work. Does anyone know why？

[torealise]

is it a Flux? Flux sometimes gives blurred images. In this case it is not a lora's issue