Weights for CoaT (vision transformer) models

Weights for CoaT: Co-Scale Conv-Attentional Image Transformers (from https://github.com/mlpc-ucsd/CoaT)