MVP: A Mobile 3D-Stacked VLM Accelerator for Efficient Video Understanding by Leveraging Dynamic Sparse Attention Patterns
MVP, a 3D-stacked VLM accelerator featuring context-aware sparse attention (CASA) and online workload-aware hybrid parallelism scheduling, is introduced, which prunes redundant attention computation FLOPS and adaptively balances computation across hybrid bonding (HB) based many-core NoC architecture.