ScaleNet: An Efficient Approach to Expanding Vision Transformer Models

ShishuLM: A New Design for Efficient Small Language Models

Making Large Language Models Leaner with Matrix Atom Sharing