ConceptAI & ML
FlashAttention
A kernel fusion operation for attention that avoids materializing the large attention matrix in high-bandwidth memory by using tiling and recomputation.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Lecturers
1Connections
3The lectures that define it or assume it, and the concepts next to it.