BootSeer: Analyzing and Mitigating Initialization Bottlenecks in Large-Scale LLM Training

Rui Li, Xiaoyun Zhi, Jinxin Chi +14

Large Language Models (LLMs) have become a cornerstone of modern AI, driving breakthroughs in natural language processing and expanding into multimodal jobs involving images, audio…