Inference Efficiency

LLaMA: No Architecture Revolution, Yet It Became the Open-Model Foundation