Superpipeline: A Universal Approach for Reducing GPU Memory Usage in Large Models

Reza Abbasi, Sernam Lim

The rapid growth in machine learning models, especially in natural language processing and computer vision, has led to challenges when running these models on hardware with limited…