Is it correct for Hailo utilization to always be 100%? I want to evaluate the utilization of HAilo running inference using YOLO models of various sizes (s vs. x), and I am getting the same 100% utilization for both.
This utilization granularity is the whole device, and you would see smaller fractions than 100% if you will use he scheduler to run more than one process. Very similar to a single core cpu.
If you want to know the actual internal utilization, you should use the profiler tool, this would give you the best utilization.
1 Like