Paged Attention

Concept

A memory management technique that allocates KV cache memory in fixed-size blocks (pages) to prevent fragmentation and efficiently manage dynamic output lengths, improving GPU utilization.

Mentioned in 1 video

Videos Mentioning Paged Attention

Stanford CME295 Transformers & LLMs | Autumn 2026 | Lecture 5 - LLM Systems

Stanford CME295 Transformers & LLMs | Autumn 2026 | Lecture 5 - LLM Systems

Stanford Online

A memory management technique that allocates KV cache memory in fixed-size blocks (pages) to prevent fragmentation and efficiently manage dynamic output lengths, improving GPU utilization.