1
ReCoVer: Resilient LLM Pre-Training System via Fault-Tolerant Collective and Versatile Workload
新论文提出ReCoVer系统,用容错集合和灵活工作负载增强LLM预训练弹性,减少训练中断损失。
arXiv:2605.11215v2 Announce Type: replace-cross Abstract: Pre-training large language models on massive GPU clusters has made hardware faults routine …