AdaMAE: Adaptive Masking for Efficient Spatiotemporal Learning With Masked Autoencoders | Read Paper on Bytez