Training with paired dataset

[CanberkSaglam]

Hello, thanks for great work. I want to train SFT layers with own paired dataset (GT and LQ pairs) where resolutions are 256x256. I tried using dataloader scripts that you provided but it does not work. First error was about get_input function of LatentDiffusionSRTextWT class. I removed degradation pipeline in get_input function so that it take GT and LQ images as they are. But now i get an error in spade.py script. Logs are in the figure below. Can you provide me a guidance for training on paired dataset? Thanks in advance.

[Justones]

Did you fix the problem?

[KenWuqianghao]

wondering the same @IceClear

[xiaoyaolovlife]

Same Question