The main responsibility of the AUTO plugin is to provide a unified device that enables developers to code deep learning applications once and deploy them anywhere.
Other capabilities of the AUTO plugin include:
- Static device selection, which intelligently loads a network to one device or multiple devices.
- CPU acceleration to start inferencing while the target device is still loading the network.
- Model priority support for loading multiple networks to multiple devices.
The component is written in C++. If you want to contribute to the AUTO plugin, follow the common coding style rules.
In case of any questions, review and merge requests, contact the AUTO Plugin maintainer group.
The AUTO plugin follows the OpenVINO™ plugin architecture and consists of several main components:
- docs contains developer documentation for the AUTO plugin.
- current folder contains sources of the AUTO plugin.
Learn more in the OpenVINO™ Plugin Developer Guide.
The diagram below shows an overview of the components responsible for the basic inference flow:
flowchart TD
subgraph Application["Application"]
end
subgraph OpenVINO Runtime["OpenVINO Runtime"]
AUTO["AUTO Plugin"] --> CPU["CPU Plugin"]
AUTO["AUTO Plugin"] --> GPU["GPU Plugin"]
end
Application --> AUTO
style Application fill:#6c9f7f
Find more details in the AUTO Plugin architecture document.