gaunernst

/

vit_base_patch16_1024_128.audiomae_as2m

like 1

I have updated the README, you can refer to it as an example. Basically you can use timm API model.forward_features() then reshape appropriately