Code for "AVG-LLaVA: A Multimodal Large Model with Adaptive Visual Granularity" - View it on GitHub