This interface extends regular plugin interface for heterogeneous case. Not all plugins implements it. The main purpose of this interface - to register loaders and have an ability to get default settings for affinity on certain devices.
More...
|
virtual void |
SetDeviceLoader (const std::string &device, IHeteroDeviceLoader::Ptr loader) noexcept=0 |
|
virtual StatusCode |
SetAffinity (ICNNNetwork &network, const std::map< std::string, std::string > &config, ResponseDesc *resp) noexcept=0 |
|
The main goal of this function to set affinity according to the options set for the plugin implementing IHeteroInferencePlugin. This function works only if all affinity in the network are empty. More...
|
|
virtual void |
GetVersion (const Version *&versionInfo) noexcept=0 |
|
Returns plugin version information. More...
|
|
virtual void |
SetLogCallback (IErrorListener &listener) noexcept=0 |
|
Sets logging callback Logging is used to track what is going on inside. More...
|
|
virtual StatusCode |
LoadNetwork (ICNNNetwork &network, ResponseDesc *resp) noexcept=0 |
|
Loads a pre-built network with weights to the engine. In case of success the plugin will be ready to infer. More...
|
|
virtual StatusCode |
LoadNetwork (IExecutableNetwork::Ptr &ret, ICNNNetwork &network, const std::map< std::string, std::string > &config, ResponseDesc *resp) noexcept=0 |
|
Creates an executable network from a network object. User can create as many networks as they need and use them simultaneously (up to the limitation of the hardware resources) More...
|
|
virtual StatusCode |
ImportNetwork (IExecutableNetwork::Ptr &ret, const std::string &modelFileName, const std::map< std::string, std::string > &config, ResponseDesc *resp) noexcept=0 |
|
Creates an executable network from a previously exported network. More...
|
|
virtual StatusCode |
Infer (const Blob &input, Blob &result, ResponseDesc *resp) noexcept=0 |
|
Infers an image(s). Input and output dimensions depend on the topology. As an example for classification topologies use a 4D Blob as input (batch, channels, width, height) and get a 1D blob as output (scoring probability vector). To Infer a batch, use a 4D Blob as input and get a 2D blob as output in both cases the method will allocate the resulted blob. More...
|
|
virtual StatusCode |
Infer (const BlobMap &input, BlobMap &result, ResponseDesc *resp) noexcept=0 |
|
Infers tensors. Input and output dimensions depend on the topology. As an example for classification topologies use a 4D Blob as input (batch, channels, width, height) and get a 1D blob as output (scoring probability vector). To Infer a batch, use a 4D Blob as input and get a 2D blob as output in both cases the method will allocate the resulted blob. More...
|
|
virtual StatusCode |
GetPerformanceCounts (std::map< std::string, InferenceEngineProfileInfo > &perfMap, ResponseDesc *resp) const noexcept=0 |
|
Queries performance measures per layer to get feedback of what is the most time consuming layer Note: not all plugins provide meaningful data. More...
|
|
virtual StatusCode |
AddExtension (InferenceEngine::IExtensionPtr extension, InferenceEngine::ResponseDesc *resp) noexcept=0 |
|
Registers extension within the plugin. More...
|
|
virtual StatusCode |
SetConfig (const std::map< std::string, std::string > &config, ResponseDesc *resp) noexcept=0 |
|
Sets configuration for plugin, acceptable keys can be found in ie_plugin_config.hpp. More...
|
|
virtual void |
QueryNetwork (const ICNNNetwork &network, QueryNetworkResult &res) const noexcept |
|
Query plugin if it supports specified network. More...
|
|
virtual void |
QueryNetwork (const ICNNNetwork &network, const std::map< std::string, std::string > &config, QueryNetworkResult &res) const noexcept |
|
Query plugin if it supports specified network with specified configuration. More...
|
|
This interface extends regular plugin interface for heterogeneous case. Not all plugins implements it. The main purpose of this interface - to register loaders and have an ability to get default settings for affinity on certain devices.
- Deprecated:
- Use InferenceEngine::Core with HETERO mode in LoadNetwork, QueryNetwork, etc