gLLM: Global Balanced Pipeline Parallelism Systems for Distributed LLMs Serving with Token Throttling.

Tianyu Guo, Xianwei Zhang, Jiangsu Du, Zhiguang Chen, Nong Xiao, Yutong Lu

Browse the full SC paper archive.