AI on Edge
- Tom Gall
- Theodore Grey
Page Owners: Tom Gall
Related Web pages:
Edge device compares a wide variety of Cortex-A equipped hardware. These devices might be running Android, Linux and every other operating systems. Within this class of devices their capability to performance AI workloads such as inference can vary greatly. On the low end memory might be tight and no offload exists to on the high end, they can be server like with offload and plenty of resources. Indeed Edge devices since they are Cortex-A can have quite a bit in common with HPC/Server with the exception that one does not generally perform training on an edge device.
AI on Edge engineering is focused on the following frameworks:
- ArmNN
- Arm Compute Library (ACL)
- ONNX / ONNX-RT
- PyTorch / Glow
- Tensorflow / Tensorflow Lite
- Tosa
- TVM
Jira
The Linaro efforts are tracked within Jira, however this system does not capture all activity by all participants. Weekly on Wednesdays an AI Engineer sync meeting is held which coordinates activity across the various teams from the participating companies. Tracking of tasks also occurs within the various project communities which the project is engaged.