Work

Towards Energy-Efficient Edge Inference in Radio Cpns: a Mixture-of-Depths Transformer Based Tri-Parallel Distributed Approach