Lynx: Enabling Efficient MoE Inference through Dynamic Batch-Aware Expert Selection

Open in new window