Transformer Architecture

Mainstream deep learning sequence modeling structure based on self-attention mechanisms.