Pre-trained using Deep Masked Fully Occluded Autoencoder Framework (FCMAE) in convnextv2 network and terrain on ImageNet-1k?

[watertianyi]

How does convnextv2 use timm.create_model or correspond to https://huggingface.co/timm/convnextv2_large.fcmae_ft_in1k, because I want to use towhee for vector retrieval？