Is Mixup necessary for MAE fine-tuning?

[SCU-zly]

Your work is so excellent! However, when I try to fine-tuning on my downstream regression task, I found that when using Mixup method, the parameter of the function contains: num_classes=args.nb_classes. Due to the special nature of my task, I have the output target of only one class. So there's two possible solution for me:

• Set the num_classes to 1 naively.
• Disable the Mixup method.
  Besides, if I use Mixup, the loss function for my regression task should also find a proper one.
  I'm confuse to use which solution or other better solution....I'm looking forward to your answer!