<ul><li>Frenzy is a memory-aware serverless computing method for heterogeneous GPU clusters.</li><li>It automates resource allocation and scheduling for efficient training of Large Language Models (LLMs).</li><li>Key features include Memory-Aware Resources Predictor (MARP) and Heterogeneity-Aware Scheduling (HAS).</li><li>Frenzy improves memory usage prediction, reduces scheduling overhead, and enhances resource allocation.</li></ul>

Frenzy: A Memory-Aware Serverless Computing Method for Heterogeneous GPU Clusters

Discover more