Abstract: Model partitioning techniques, which decompose and collaboratively execute subtasks of deep neural networks (DNNs), have emerged as a critical strategy for enhancing distributed inference ...