Skip to visualization
←visualization / PagedAttention

minGPT · PagedAttention

paged KV-cache memory
◆ PAGED KV CACHE
T=8
i4

🔍 Variable Inspector

click any code line
Click a code line to inspect its variables.
About this lesson
Quick answer

What PagedAttention explains

Map logical KV blocks onto non-contiguous GPU pages to reduce fragmentation, batch requests efficiently, and share cached prefixes.

Key concepts

Primary research

vLLM · Kwon et al. 2023