🧩 Attention KernelsSpecificFused Attention, Tiling Strategies, Memory Hierarchy, Transformer Optimization