NEW StreamingLLM by MIT & Meta: Code explained

MIT and META introduce StreamingLLM, an efficient framework
that enables LLMs trained with a finite length attention window to generalize to
infinite sequence length without any fine-tuning. Streaming LLM.

ARXIV preprint:
https://arxiv.org/pdf/2309.17453v1.pdf

GitHub repo:
https://github.com/mit-han-lab/streaming-llm/blob/main/streaming_llm/pos_shift/modify_llama.py

arxiv attention code explained fine-tuning framework github github repo llm llms meta mit streaming