Efficiently accelerating federated learning on resource-constrained edge devices through adaptive model partitioning and bandwidth allocation.