2、《
ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders