LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models

Open in new window