Comparison of Autoencoders for tokenization of ASL datasets