The design of the original Transformer model in the Attention is all You Need paper was predicated on efficiency (all layers the same size, combining word/token

The design of the original Transformer model in the Attention is all You Need paper was predicated on efficiency (all layers the same size, combining word/token embedding with position in the input stream harmonic embedding). It is good to see improvements!