a6351294e7
* Change `VPUX` occurences to `NPU` * Change library for `NPU` device in `api_conformance_helpers.hpp` * Rename `MYRIAD plugin` * Switch `HARDWARE_AWARE_IGNORED_PATTERNS` VPU to NPU * Rename DEVICE_KEEMBAY to DEVICE_NPU * Rename VPUX_DEVICE_NAME to NPU_DEVICE_NAME * Rename vpu_patterns to npu_patterns * Change VPUX occurences to NPU after review * Remove VPUX device comment * Change VPUX/vpu to NPU in tests/time_tests * Rename VPU to NPU in docs after review * Rename VPU to NPU in tools/pot after review * Renamed vpu.json to npu.json in tools/pot after review * Restore CommonTestUtils::DEVICE_KEEMBAY --------- Co-authored-by: MirceaDan99 <mircea-aurelian.dan@intel.com>
OpenVINO™ AUTO Plugin
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.
Key contacts
In case of any questions, review and merge requests, contact the AUTO Plugin maintainer group.
Components
The AUTO plugin follows the OpenVINO™ plugin architecture and consists of several main components:
- docs contains developer documentation for the AUTO plugin.
- src - folder contains sources of the AUTO plugin.
- tests - tests for Auto Plugin components.
Learn more in the OpenVINO™ Plugin Developer Guide.
Architecture
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.