KV Cache, Flash Attention & Inference Optimization
Training is parallel and FLOP-bound. Inference is serial and memory-bound. Different bottleneck, different tricks.
Visual edition planned
This lesson isn’t interactive yet.
It is part of the curriculum and will get the same treatment as Phase 1 — a visual cover, hands-on labs, derivations with numeric checks and a quiz. Until then, the original lesson is the best place to read it:
Part of Phase 07 — Transformers Deep Dive. Use the previous / next cards below to keep browsing the phase.