ResearchHugging Face

Same Cluster, 33 Points More Utilization: What Changed Was the Order

#gpu#utilization#scheduling#ai#performance

English

Hugging Face developed a constraint-aware GPU allocator that improved GPU utilization by up to 33 percentage points compared to a FIFO scheduler. The key factor was the order of allocation decisions, which allowed for better handling of competing workload types, ultimately enhancing priority-weighted output by as much as 105%.

中文

Hugging Face 开发了一种约束感知的 GPU 分配器，与 FIFO 调度器相比，GPU 利用率提高了最多 33 个百分点。关键因素在于分配决策的顺序，使得更好地处理竞争工作负载类型，最终优先权加权输出提高了多达 105%。