Currently the Encoder and Decoder Layer are are Quaternion-NN. Prediction Layer is Real fully connected layer. Embeddig Layer is also Real.
*Testing code will be added shortly.
**Implementation of Quaternion Transformer is best to my understanding. Any improvement in implementation is appreciated.