Learn how to build and optimize attention mechanisms for transformer models, from basic self-attention to the multi-head attention architecture used in state-of-the-art language models