作者位次 #2/9OSDI 2026Oral· Operating Systems & Distributed Systems
OpGuard: Bitwise Alignment for Precise and General Debugging of Production LLM Training
Ziming Zhou · Yinjie Zhao · Hang Zhu · Wenxiao Wang · Zhihao Bai · Yun Zhang · Shuguang Wang · Haibin Lin · Peng Huang
摘要
Large-scale LLM training runs on many GPUs for weeks atop rapidly evolving software stacks. Bugs or hardware glitches can silently corrupt the computation and only surface much later. Debugging becomes finding a needle in a haystack across time. Developers often use another training run and compa...