Flash Attention

An efficient attention algorithm reducing memory usage and increasing speed through clever recomputation.

This concept is essential for understanding training & optimization and forms a key part of modern AI systems.