Nashville AI/ML Reading Group (Paged Attention)

Hi everyone, we’re meeting today, October 9th at 7 PM CST! This week, we’ll be discussing Paged Attention and an introduction to using VLLM.Looking forward to seeing you all there!

Meeting link: https://meet.google.com/vqh-fnge-arc

📄 Abstract

High throughput serving of large language models (LLMs) requires batching sufficiently many requests at a time. However, existing systems struggle because the key-value cache (KV cache) memory for each request is huge and grows and shrinks dynamically. When managed inefficiently, this memory can be significantly wasted by fragmentation and redundant duplication, limiting the batch size. To address this problem, we propose PagedAttention, an attention algorithm inspired by the classical virtual memory and paging techniques in operating systems. On top of it, we build vLLM, an LLM serving system that achieves (1) near-zero waste in KV cache memory and (2) flexible sharing of KV cache within and across requests to further reduce memory usage. Our evaluations show that vLLM improves the throughput of popular LLMs by 2-4× with the same level of latency compared to the state-of-the-art systems, such as FasterTransformer and Orca. The improvement is more pronounced with longer sequences, larger models, and more complex decoding algorithms.

📄 Paper in review

Title: Efficient Memory Management for Large Language Model Serving with PagedAttention

Authors: Woosuk Kwon, et al.
Link: https://openreview.net/forum?id=DO2WFXU1Be

Presenter: Alexy Skoutnev

See you Today!

Best regards,
Alexy Skoutnev
Organizer, Nashville AI/ML Reading Group