ie_icnn_network.hpp

Go to the documentation of this file.

 // Copyright (C) 2018-2019 Intel Corporation


 // SPDX-License-Identifier: Apache-2.0


 //


 


 /**


  * @brief This is a header file for the ICNNNetwork class


  * @file ie_icnn_network.hpp


  */


 #pragma once


 

 #include "ie_common.h"


 #include "ie_layers.h"


 #include "ie_data.h"


 #include "ie_device.hpp"


 #include "ie_blob.h"


 #include "details/ie_irelease.hpp"


 #include "ie_preprocess.hpp"


 #include "ie_input_info.hpp"


 #include "ie_icnn_network_stats.hpp"


 #include "ie_iextension.h"


 #include <memory>


 #include <map>


 #include <string>


 

 namespace InferenceEngine {

 


 /**


  * @brief A collection that contains string as key, and Data smart pointer as value


  */


 using OutputsDataMap = std::map<std::string, DataPtr>;

 


 /**


  * @brief This is the main interface to describe the NN topology


  */


 class ICNNNetwork : public details::IRelease {

 public:

     using Ptr = std::shared_ptr<ICNNNetwork>;

 


     /**


      * @brief Returns the main network operating precision.


      * This may be MIXED if not homogeneous.


      * @return A precision type


      */


     virtual Precision getPrecision() const noexcept = 0;

 


     /**


      * @brief Gets the network output Data node information. The received info is stored in the given Data node.


      * For single and multiple outputs networks.


      * @param out Reference to the OutputsDataMap object


      */


     virtual void getOutputsInfo(OutputsDataMap& out) const noexcept  = 0;

 


     /**


      * @brief Gets the network input Data node information. The received info is stored in the given InputsDataMap object.


      * For single and multiple inputs networks.


      * This method must be called to find out input names for using them later during filling of a map


      * of blobs passed later to InferenceEngine::IInferencePlugin::Infer()


      * @param inputs Reference to InputsDataMap object.


      */


     virtual void getInputsInfo(InputsDataMap& inputs) const noexcept  = 0;

 

 


     /**


      * @brief Returns information on certain input pointed by inputName


      * @param inputName Name of input layer to get info on


      * @return A smart pointer to the input information


      */


     virtual InputInfo::Ptr getInput(const std::string& inputName) const noexcept = 0;

 


     /**


      * @brief Gets the network name. The name is stored in the given pName string.


      * @param pName - will receive actual network name, specified in IR file,


      *     pName should point to valid memory address before invoking this function


      * @param len - size in bytes of pName buffer, actual name is trimmed by this size


      */


     virtual void getName(char* pName, size_t len) const noexcept = 0;

 


     /**


      * @brief Returns the network name.


      * @return Network name


      */


     virtual const std::string& getName() const noexcept = 0;

 


     /**


     * @brief Returns the number of layers in the network as an integer value


     * @return The number of layers as an integer value


     */


     virtual size_t layerCount() const noexcept = 0;

 


     /**


     * @brief Returns a smart pointer reference to a Data node given its name.


      * If the Data node is missing, returns reference to a default initialized new empty data pointer with given name.


     * @param dname Name of the Data node


     * @return Data node smart pointer


     */


     virtual DataPtr& getData(const char* dname)  noexcept = 0;

 


     /**


     * @brief Insert a layer into the network. A user is responsible to connect it to other data elements.


     * @param layer Const reference to a layer smart pointer


     */


     virtual void addLayer(const CNNLayerPtr& layer) noexcept = 0;

 


     /**


      * @brief Adds output to the layer


      * @param layerName Name of the layer


      * @param outputIndex Index of the output


      * @param resp Response message


      * @return Status code of the operation


      */


     virtual StatusCode


     addOutput(const std::string& layerName, size_t outputIndex = 0, ResponseDesc* resp = nullptr) noexcept = 0;

 


     /**


      * @brief Gets network layer with the given name


      * @param layerName Given name of the layer


      * @param out Pointer to the found CNNLayer object with the given name


      * @param resp Pointer to the response message that holds a description of an error if any occurred


      * @return Status code of the operation. OK if succeeded


      */


     virtual StatusCode getLayerByName(const char* layerName, CNNLayerPtr& out, ResponseDesc* resp) const noexcept = 0;

 


     /**


      * @brief Sets a desirable device to perform all work on.


      * Some plug-ins might not support some target devices and may abort execution with an appropriate error message.


      * @param device Device to set as a target


      */


     virtual void setTargetDevice(TargetDevice device) noexcept = 0;

 


     /**


      * @brief Gets the target device.


      * If setTargetDevice() was not called before, returns eDefault


      * @return A TargetDevice instance


      */


     virtual TargetDevice getTargetDevice() const noexcept = 0;

 


     /**


      * @deprecated use setBatchSize with ResponseDesc to get error message


      * @brief Changes the inference batch size


      */


     virtual StatusCode setBatchSize(const size_t size) noexcept = 0;

 


     /**


      * @brief Changes the inference batch size.


      * @note There are several limitations and it's not recommended to use it. Set batch to the input shape and call @reshape.


      * @param size Size of batch to set


      * @return Status code of the operation


      * @note: Current implementation of the function sets batch size to the first dimension of all layers in the networks.


      * Before calling it make sure that all your layers have batch in the first dimension, otherwise the method works incorrectly.


      * This limitation is resolved via shape inference feature


      * by using InferenceEngine::ICNNNetwork::reshape method.


      * To read more refer to the Shape Inference section in documentation


      */


     virtual StatusCode setBatchSize(size_t size, ResponseDesc* responseDesc) noexcept = 0;

 


     /**


     * @brief Gets the inference batch size


     * @return The size of batch as a size_t value


     */


     virtual size_t getBatchSize() const noexcept = 0;

 


     /**


      * @brief Map of pairs: name of corresponding data and its dimension.


      */


     using InputShapes = std::map<std::string, SizeVector>;

 


     /**


      * @brief Run shape inference with new input shapes for the network


      * @param inputShapes - map of pairs: name of corresponding data and its dimension.


      * @param resp Pointer to the response message that holds a description of an error if any occurred


      * @return Status code of the operation


      */


     virtual StatusCode reshape(const InputShapes& /*inputShapes*/, ResponseDesc* /*resp*/) noexcept { return NOT_IMPLEMENTED; };

 


     /**


      * @brief Registers extension within the plugin


      * @param extension Pointer to already loaded reader extension with shape propagation implementations


      * @param resp Pointer to the response message that holds a description of an error if any occurred


      * @return Status code of the operation. OK if succeeded


      */


     virtual StatusCode


     AddExtension(const IShapeInferExtensionPtr& /*extension*/, ResponseDesc* /*resp*/) noexcept { return NOT_IMPLEMENTED; };

 

     virtual StatusCode getStats(ICNNNetworkStats** /*stats*/, ResponseDesc* /*resp*/) const noexcept { return NOT_IMPLEMENTED; };

 


     /**


      * @brief Serialize network to IR and weights files.


      * @param xmlPath Path to output IR file.


      * @param binPath Path to output weights file.


      * @return Status code of the operation


      */


     virtual StatusCode serialize(const std::string &xmlPath, const std::string &binPath, ResponseDesc* resp) const noexcept = 0;

 };

 }  // namespace InferenceEngine