LLM in a Flash: Efficient Large Language Model Inference with Limited Memory

Previous discussion here: https://news.ycombinator.com/item?id=38704982

Links are different though.

This link is to arxiv. The one in the discussion I link is to some hugging face papers reference.