apq
apq copied to clipboard
Changning latency constraint seems to have little to no effect on the results
When running search.py
with pre-trained weights provided in the README I am getting consistent accuracy results independently of the latency constraint (120ms yields similar results to 8ms). How could that be?