Executable Network

ExecutableNetwork class functionality:

• Compile an InferenceEngine::ICNNNetwork instance to a backend specific graph representation
• Create an arbitrary number of InferRequest objects
• Hold some common resources shared between different instances of InferRequest. For example:
  • InferenceEngine::IExecutableNetworkInternal::_taskExecutor task executor to implement asynchronous execution
  • InferenceEngine::IExecutableNetworkInternal::_callbackExecutor task executor to run an asynchronous inference request callback in a separate thread

ExecutableNetwork Class

Inference Engine Plugin API provides the helper InferenceEngine::ExecutableNetworkThreadSafeDefault class recommended to use as a base class for an executable network. Based on that, a declaration of an executable network class can look as follows:

class ExecutableNetwork : public InferenceEngine::ExecutableNetworkThreadSafeDefault {
public:
    ExecutableNetwork(const std::shared_ptr<const ngraph::Function>& function, const InferenceEngine::InputsDataMap& inputInfoMap,
                      const InferenceEngine::OutputsDataMap& outputsInfoMap, const Configuration& cfg, const std::shared_ptr<Plugin>& plugin);
 
    ExecutableNetwork(std::istream& model, const Configuration& cfg, const std::shared_ptr<Plugin>& plugin);
 
    // Methods from a base class ExecutableNetworkThreadSafeDefault
 
    void Export(std::ostream& model) override;
    InferenceEngine::IInferRequestInternal::Ptr CreateInferRequestImpl(InferenceEngine::InputsDataMap networkInputs,
                                                                       InferenceEngine::OutputsDataMap networkOutputs) override;
    InferenceEngine::IInferRequestInternal::Ptr CreateInferRequest() override;
    InferenceEngine::Parameter GetMetric(const std::string& name) const override;
    InferenceEngine::Parameter GetConfig(const std::string& name) const override;
 
private:
    friend class TemplateInferRequest;
 
    void CompileNetwork(const std::shared_ptr<const ngraph::Function>& function, const InferenceEngine::InputsDataMap& inputInfoMap,
                        const InferenceEngine::OutputsDataMap& outputsInfoMap);
    void InitExecutor();
 
    std::atomic<std::size_t> _requestId = {0};
    Configuration _cfg;
    std::shared_ptr<Plugin> _plugin;
    std::shared_ptr<ngraph::Function> _function;
    std::map<std::string, std::size_t> _inputIndex;
    std::map<std::string, std::size_t> _outputIndex;
};

Class Fields

The example class has several fields:

• _requestId - Tracks a number of created inference requests, which is used to distinguish different inference requests during profiling via the Intel® Instrumentation and Tracing Technology (ITT) library.
• _cfg - Defines a configuration an executable network was compiled with.
• _plugin - Refers to a plugin instance.
• _function - Keeps a reference to transformed ngraph::Function which is used in ngraph reference backend computations. Note, in case of other backends with backend specific graph representation _function has different type and represents backend specific graph or just a set of computational kernels to perform an inference.
• _inputIndex - maps a name of input with its index among all network inputs.
• _outputIndex - maps a name of output with its index among all network outputs.

ExecutableNetwork Constructor with ICNNNetwork

This constructor accepts a generic representation of a neural network as an InferenceEngine::ICNNNetwork reference and is compiled into a backend specific device graph:

TemplatePlugin::ExecutableNetwork::ExecutableNetwork(const std::shared_ptr<const ngraph::Function>& function,
                                                     const InferenceEngine::InputsDataMap& inputInfoMap, const InferenceEngine::OutputsDataMap& outputsInfoMap,
                                                     const Configuration& cfg, const Plugin::Ptr& plugin)
    : InferenceEngine::ExecutableNetworkThreadSafeDefault(nullptr, nullptr),  // Disable default threads creation
      _cfg(cfg),
      _plugin(plugin) {
    // TODO: if your plugin supports device ID (more that single instance of device can be on host machine)
    // you should select proper device based on KEY_DEVICE_ID or automatic behavior
    // In this case, _waitExecutor should also be created per device.
    try {
        CompileNetwork(function, inputInfoMap, outputsInfoMap);
        InitExecutor();  // creates thread-based executor using for async requests
    } catch (const InferenceEngine::Exception&) {
        throw;
    } catch (const std::exception& e) {
        IE_THROW(Unexpected) << "Standard exception from compilation library: " << e.what();
    } catch (...) {
        IE_THROW(Unexpected) << "Generic exception is thrown";
    }
}

The implementation CompileNetwork is fully device-specific.

CompileNetwork()

The function accepts a const shared pointer to ngraph::Function object and performs the following steps:

1. Applies ngraph passes using TransformNetwork function, which defines plugin-specific conversion pipeline.
2. Maps the transformed graph to a backend specific graph representation (for example, to MKLDNN graph for Intel CPU).
3. Allocates and fills memory for graph weights, backend specific memory handles and so on.

// forward declaration
std::shared_ptr<ngraph::Function> TransformNetwork(const std::shared_ptr<const ngraph::Function>& function, const InferenceEngine::InputsDataMap& inputInfoMap,
                                                   const InferenceEngine::OutputsDataMap& outputsInfoMap);
 
void TemplatePlugin::ExecutableNetwork::CompileNetwork(const std::shared_ptr<const ngraph::Function>& function,
                                                       const InferenceEngine::InputsDataMap& inputInfoMap,
                                                       const InferenceEngine::OutputsDataMap& outputsInfoMap) {
    // TODO: perform actual graph compilation / mapping to backend graph representation / kernels
 
    // apply plugins transformations
    _function = TransformNetwork(function, inputInfoMap, outputsInfoMap);
 
    // Generate backend specific blob mappings. For example Inference Engine uses not ngraph::Result nodes friendly name
    // as inference request output names but the name of the layer before.
    for (auto&& result : _function->get_results()) {
        auto previousOutput = result->get_input_source_output(0);
        auto outputName = previousOutput.get_node()->get_friendly_name();
        if (previousOutput.get_node()->get_output_size() > 1) {
            outputName += '.' + std::to_string(previousOutput.get_index());
        }
        _outputIndex.emplace(outputName, _function->get_result_index(result));
    }
    for (auto&& parameter : _function->get_parameters()) {
        _inputIndex.emplace(parameter->get_friendly_name(), _function->get_parameter_index(parameter));
    }
 
    // Perform any other steps like allocation and filling backend specific memory handles and so on
}

NOTE: After all these steps, the backend specific graph is ready to create inference requests and perform inference.

ExecutableNetwork Constructor Importing from Stream

This constructor creates a backend specific graph by importing from a stream object:

NOTE: The export of backend specific graph is done in the Export method, and data formats must be the same for both import and export.

TemplatePlugin::ExecutableNetwork::ExecutableNetwork(std::istream& model, const Configuration& cfg, const Plugin::Ptr& plugin): _cfg(cfg), _plugin(plugin) {
    // read XML content
    std::string xmlString;
    std::uint64_t dataSize = 0;
    model.read(reinterpret_cast<char*>(&dataSize), sizeof(dataSize));
    xmlString.resize(dataSize);
    model.read(const_cast<char*>(xmlString.c_str()), dataSize);
 
    // read blob content
    InferenceEngine::Blob::Ptr dataBlob;
    model.read(reinterpret_cast<char*>(&dataSize), sizeof(dataSize));
    if (0 != dataSize) {
        dataBlob = InferenceEngine::make_shared_blob<std::uint8_t>(
            InferenceEngine::TensorDesc(InferenceEngine::Precision::U8, {static_cast<std::size_t>(dataSize)}, InferenceEngine::Layout::C));
        dataBlob->allocate();
        model.read(dataBlob->buffer(), dataSize);
    }
 
    // TODO: implement Import / Export of configuration options and merge with `cfg`
    // TODO: implement Import / Export of network precisions, layouts, preprocessing info
    InferenceEngine::InputsDataMap inputInfoMap;
    InferenceEngine::OutputsDataMap outputInfoMap;
 
    auto cnnnetwork = _plugin->GetCore()->ReadNetwork(xmlString, std::move(dataBlob));
 
    setNetworkInputs(cnnnetwork.getInputsInfo());
    setNetworkOutputs(cnnnetwork.getOutputsInfo());
    SetPointerToPlugin(_plugin->shared_from_this());
 
    try {
        CompileNetwork(cnnnetwork.getFunction(), inputInfoMap, outputInfoMap);
        InitExecutor();  // creates thread-based executor using for async requests
    } catch (const InferenceEngine::Exception&) {
        throw;
    } catch (const std::exception& e) {
        IE_THROW(Unexpected) << "Standard exception from compilation library: " << e.what();
    } catch (...) {
        IE_THROW(Unexpected) << "Generic exception is thrown";
    }
}

Export()

The implementation of the method should write all data to the model stream, which is required to import a backend specific graph later in the Plugin::Import method:

void TemplatePlugin::ExecutableNetwork::Export(std::ostream& modelStream) {
    OV_ITT_SCOPED_TASK(itt::domains::TemplatePlugin, "ExecutableNetwork::Export");
 
    // Note: custom ngraph extensions are not supported
    std::map<std::string, ngraph::OpSet> custom_opsets;
    std::stringstream xmlFile, binFile;
    ngraph::pass::Serialize serializer(xmlFile, binFile, ngraph::pass::Serialize::Version::IR_V10, custom_opsets);
    serializer.run_on_function(_function);
 
    auto m_constants = binFile.str();
    auto m_model = xmlFile.str();
 
    auto dataSize = static_cast<std::uint64_t>(m_model.size());
    modelStream.write(reinterpret_cast<char*>(&dataSize), sizeof(dataSize));
    modelStream.write(m_model.c_str(), dataSize);
 
    dataSize = static_cast<std::uint64_t>(m_constants.size());
    modelStream.write(reinterpret_cast<char*>(&dataSize), sizeof(dataSize));
    modelStream.write(reinterpret_cast<char*>(&m_constants[0]), dataSize);
 
    // TODO: implement network precision, layout, preprocessing info serialization
}

CreateInferRequest()

The method creates an asynchronous inference request and returns it. While the public Inference Engine API has a single interface for inference request, which can be executed in synchronous and asynchronous modes, a plugin library implementation has two separate classes:

• Synchronous inference request, which defines pipeline stages and runs them synchronously in the Infer method.
• Asynchronous inference request, which is a wrapper for a synchronous inference request and can run a pipeline asynchronously. Depending on a device pipeline structure, it can has one or several stages:
  • For single-stage pipelines, there is no need to define this method and create a class derived from InferenceEngine::AsyncInferRequestThreadSafeDefault. For single stage pipelines, a default implementation of this method creates InferenceEngine::AsyncInferRequestThreadSafeDefault wrapping a synchronous inference request and runs it asynchronously in the _taskExecutor executor.
  • For pipelines with multiple stages, such as performing some preprocessing on host, uploading input data to a device, running inference on a device, or downloading and postprocessing output data, schedule stages on several task executors to achieve better device use and performance. You can do it by creating a sufficient number of inference requests running in parallel. In this case, device stages of different inference requests are overlapped with preprocessing and postprocessing stage giving better performance.

    IMPORTANT: It is up to you to decide how many task executors you need to optimally execute a device pipeline.

    InferenceEngine::IInferRequestInternal::Ptr TemplatePlugin::ExecutableNetwork::CreateInferRequest() {
        auto internalRequest = CreateInferRequestImpl(_networkInputs, _networkOutputs);
        return std::make_shared<TemplateAsyncInferRequest>(std::static_pointer_cast<TemplateInferRequest>(internalRequest), _taskExecutor, _plugin->_waitExecutor,
                                                           _callbackExecutor);
    }

    CreateInferRequestImpl()

This is a helper method used by CreateInferRequest to create a synchronous inference request, which is later wrapped with the asynchronous inference request class:

InferenceEngine::IInferRequestInternal::Ptr TemplatePlugin::ExecutableNetwork::CreateInferRequestImpl(InferenceEngine::InputsDataMap networkInputs,
                                                                                                      InferenceEngine::OutputsDataMap networkOutputs) {
    return std::make_shared<TemplateInferRequest>(networkInputs, networkOutputs, std::static_pointer_cast<ExecutableNetwork>(shared_from_this()));
}

GetMetric()

Returns a metric value for a metric with the name name. A metric is a static type of information about an executable network. Examples of metrics:

• EXEC_NETWORK_METRIC_KEY(NETWORK_NAME) - name of an executable network
• EXEC_NETWORK_METRIC_KEY(OPTIMAL_NUMBER_OF_INFER_REQUESTS) - heuristic to denote an optimal (or at least sub-optimal) number of inference requests needed to run asynchronously to use the current device fully
• Any other executable network metric specific for a particular device. Such metrics and possible values must be declared in a plugin configuration public header, for example, template/template_config.hpp

InferenceEngine::Parameter TemplatePlugin::ExecutableNetwork::GetMetric(const std::string& name) const {
    // TODO: return more supported values for metrics
    if (EXEC_NETWORK_METRIC_KEY(SUPPORTED_METRICS) == name) {
        IE_SET_METRIC_RETURN(SUPPORTED_METRICS, std::vector<std::string> {METRIC_KEY(NETWORK_NAME), METRIC_KEY(SUPPORTED_METRICS),
                                                                          METRIC_KEY(SUPPORTED_CONFIG_KEYS), METRIC_KEY(OPTIMAL_NUMBER_OF_INFER_REQUESTS)});
    } else if (EXEC_NETWORK_METRIC_KEY(SUPPORTED_CONFIG_KEYS) == name) {
        std::vector<std::string> configKeys = {CONFIG_KEY(DEVICE_ID), CONFIG_KEY(PERF_COUNT), TEMPLATE_CONFIG_KEY(THROUGHPUT_STREAMS)};
        auto streamExecutorConfigKeys = InferenceEngine::IStreamsExecutor::Config {}.SupportedKeys();
        for (auto&& configKey : streamExecutorConfigKeys) {
            configKeys.emplace_back(configKey);
        }
        IE_SET_METRIC_RETURN(SUPPORTED_CONFIG_KEYS, configKeys);
    } else if (EXEC_NETWORK_METRIC_KEY(NETWORK_NAME) == name) {
        auto networkName = _function->get_friendly_name();
        IE_SET_METRIC_RETURN(NETWORK_NAME, networkName);
    } else if (EXEC_NETWORK_METRIC_KEY(OPTIMAL_NUMBER_OF_INFER_REQUESTS) == name) {
        unsigned int value = _cfg._streamsExecutorConfig._streams;
        IE_SET_METRIC_RETURN(OPTIMAL_NUMBER_OF_INFER_REQUESTS, value);
    } else {
        IE_THROW() << "Unsupported ExecutableNetwork metric: " << name;
    }
}

The IE_SET_METRIC_RETURN helper macro sets metric value and checks that the actual metric type matches a type of the specified value.

GetConfig()

Returns a current value for a configuration key with the name name. The method extracts configuration values an executable network is compiled with.

InferenceEngine::Parameter TemplatePlugin::ExecutableNetwork::GetConfig(const std::string& name) const {
    return _cfg.Get(name);
}

This function is the only way to get configuration values when a network is imported and compiled by other developers and tools (for example, the Compile tool).

The next step in plugin library implementation is the Synchronous Inference Request class.