Implement a test where the client runs the local inference with the validation images. A certain performance is obtained, and it must be compared with the performance where the client executes up to the k-th layer and then offloads. The two performances must be similar, around 2-3% of difference at most.
Implement a test where the client runs the local inference with the validation images. A certain performance is obtained, and it must be compared with the performance where the client executes up to the k-th layer and then offloads. The two performances must be similar, around 2-3% of difference at most.