Saturday, December 14, 2024
A high-throughput and memory-efficient inference and serving engine for LLMs