This class is used to compile the model for the inference engine plugin;. More...

#include <CelanturSDKInterface.h>

Public Member Functions
	ModelCompiler (std::filesystem::path license, ModelCompilerParams params)

	ModelCompiler (const ModelCompiler &other)

ModelCompiler &	operator= (const ModelCompiler &other)

	ModelCompiler (ModelCompiler &&other) noexcept

ModelCompiler &	operator= (ModelCompiler &&other) noexcept

	~ModelCompiler ()

celantur::InferenceEnginePluginCompileSettings	preload_model (std::filesystem::path model_path)
	Preload the model from the given path and return the settings that are needed to compile the model;. More...

void	compile_model (celantur::InferenceEnginePluginCompileSettings settings, std::filesystem::path output_path)
	Compile the preloaded model with the given settings and save it to the given output path;. More...

Detailed Description

This class is used to compile the model for the inference engine plugin;.

If your inference engine plugin requires model compilation, like TensorRT or OpenVINO, you need to use this class to compile the model before using it in the Processor class;

Constructor & Destructor Documentation

◆ ModelCompiler() [1/3]

CelanturSDK::ModelCompiler::ModelCompiler	(	std::filesystem::path	license,
		ModelCompilerParams	params
	)

References CelanturSDK::ModelCompilerParams::inference_plugin.

◆ ModelCompiler() [2/3]

CelanturSDK::ModelCompiler::ModelCompiler ( const ModelCompiler & other )

◆ ModelCompiler() [3/3]

CelanturSDK::ModelCompiler::ModelCompiler ( CelanturSDK::ModelCompiler && other )

noexcept

◆ ~ModelCompiler()

CelanturSDK::ModelCompiler::~ModelCompiler ( )

Member Function Documentation

◆ compile_model()

void CelanturSDK::ModelCompiler::compile_model	(	celantur::InferenceEnginePluginCompileSettings	settings,
		std::filesystem::path	output_path
	)

Compile the preloaded model with the given settings and save it to the given output path;.

Parameters

settings	The settings that are needed to compile the model; these settings are obtained from the preload_model() function.
output_path	The path to save the compiled model; the path should include the file name and extension.

The output path should be a valid path to save the compiled model; we recommend file extension based on the inference engine plugin used:

TensorRT: .trt.enc
OpenVINO: .ovino.enc
ONNX: .onnx.enc But it is not required; you can use any extension you want. The tests we use test for these extensions.

◆ operator=() [1/2]

CelanturSDK::ModelCompiler & CelanturSDK::ModelCompiler::operator= ( const ModelCompiler & other )

◆ operator=() [2/2]

CelanturSDK::ModelCompiler & CelanturSDK::ModelCompiler::operator= ( CelanturSDK::ModelCompiler && other )

noexcept

◆ preload_model()

celantur::InferenceEnginePluginCompileSettings CelanturSDK::ModelCompiler::preload_model ( std::filesystem::path model_path )

Preload the model from the given path and return the settings that are needed to compile the model;.

The preloading is needed to extract the model specific settings that are needed to compile the model for the given inference engine plugin; it is impossible to determine the settings purely from the inference engine For more abour inference engines and their settings, please read the documentation: Inference Engines