mit-han-lab/streaming-llm

[ICLR 2024] Efficient Streaming Language Models with Attention Sinks

421 best points · 65 comments · 1 Hacker News post · 7,276 GitHub stars · View on GitHub

Hacker News discussions

• Efficient streaming language models with attention sinks (421 points, 65 comments, by guywithabowtie, 2023-10-02)