LeVit, Visformer, RedNet/Involution models and layers

[rwightman]

No description provided.

[amaarora]

@rwightman
Commenting here again as this is a better place for it.

You should be able to find visformer_small weights here - https://www.kaggle.com/aroraaman/visformer-small-chkpt

Val accuracy - 82.1%, paper reports - 82.28%. Val img size - 224x224.

Please note that I didn't train the model myself but copied the weights from here and edited the keys so that the pretrained weights can be loaded to TIMM.

Hope this helps :)

[LvJC]

Where's the RedNet pretrained weights?
I load rednet50ts but get a warning:

No pretrained weights exist for this model. Using random initialization.

[rwightman]

@LvJC clearly they don't exist, this is a development branch