3 ms·LLM in a Flash: Efficient Large Language Model Inference with Limited Memory2 points by mpweiher 3y ago