Appears that this is an open source implementation of the same "Efficient Streaming Language Models with Attention Sinks" paper from MIT, linked here 7 days ago. Published on Sept 29, 2023.

https://news.ycombinator.com/item?id=37740932