Hello! Could you provide the training config for DiTDH-B(DINOv2-B)? Additionally, I’ve noticed that the online evaluation FID is very high, whereas the offline evaluation FID remains normal. Is this discrepancy caused by the two evaluation methods using different libraries?
Hello! Could you provide the training config for DiTDH-B(DINOv2-B)? Additionally, I’ve noticed that the online evaluation FID is very high, whereas the offline evaluation FID remains normal. Is this discrepancy caused by the two evaluation methods using different libraries?