Inference without Interference: Disaggregate LLM Inference for Mixed Downstream Workloads

Hu, Cunchen, Huang, Heyang, Xu, Liangliang · arXiv · 2024