Sequence-to-Sequence Models

This note is a placeholder for encoder-decoder design, attention-based alignment, and training/decoding strategies.