Search papers, labs, and topics across Lattice.
Aarhus University
1
0
3
Naive distributed inference on edge devices can be *slower* than local execution due to CPU-GPU communication bottlenecks, but a profiling-driven adaptive approach can flip the script for significant gains.
