Description
Neural-network workloads can test the development compiler path for supported Intel NPU hardware.
It is useful for developers tracking current upstream behavior for local AI acceleration. Git builds can change model support and generated output, so keep stable toolchains for production deployments.