Systems and efficient execution · MLD-0869
PagedAttention
An attention-memory approach that manages key-value caches in blocks rather than requiring contiguous allocation per request.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Serve requests with less wasted cache memory.
Original caveat
It addresses cache management rather than changing the learned attention objective.
Context and aliases
Level: Advanced. Stages: Training · Inference · Deployment.
Tags: DL · LLM · SLM · Systems.
Aliases: None recorded.
Related concepts
Concepts linking here
Original references
- PyTorch Fully Sharded Data Parallel ↗ · Consulted
- PyTorch documentation ↗ · Reference link
- vLLM documentation ↗ · Consulted
- Original reference URL ↗