Tsinghua University

This research investigates techniques for making the training of large language models more environmentally sustainable without compromising model performance. The authors propose novel methods for reducing energy consumption during training, including adaptive batch sizing, efficient model architectures, and intelligent resource allocation. The study provides extensive empirical analysis of different training strategies and their impact on both model quality and environmental footprint.

This comprehensive study explores methods for developing energy-efficient vision transformers while maintaining high performance in computer vision tasks. The research evaluates various optimization techniques including architecture modifications, training strategies, and inference optimizations specifically designed for vision transformers. The authors demonstrate significant reductions in computational costs and energy consumption while preserving model accuracy across different vision tasks.