This is a cache of https://discuss.96boards.org/t/executing-inference-workloads-on-npu-hikey-970/10882. It is a snapshot of the page at 2021-06-06T03:40:38.502+0000.
Executing inference workloads on NPU( <strong>hikey</strong> 970) - <strong>hikey</strong> - 96Boards Forum

Executing inference workloads on NPU( hikey 970)

Hello.
I have installed Debian as an OS in the hikey970 board. How can neural network workload be executed on NPU? Is there any specific API functions? And, I have read that NPU has memory constraint and will not execute if size exceeds 100MB. How can we specify computing device as NPU for certain layers ?
Thanks.