UE::NNE::IModelGPU Class Reference

#include <NNERuntimeGPU.h>

Public Member Functions
virtual	~IModelGPU ()=default
virtual TSharedPtr< UE::NNE::IModelInstanceGPU >	CreateModelInstanceGPU ()=0

Detailed Description

The interface of a model capable of creating model instance that can run on GPU.

Use UE::NNE::GetRuntime<INNERuntimeGPU>(RuntimeName) to get a runtime capable of creating GPU models.

Constructor & Destructor Documentation

~IModelGPU()

virtual UE::NNE::IModelGPU::~IModelGPU ( )

virtualdefault

Member Function Documentation

CreateModelInstanceGPU()

virtual TSharedPtr< UE::NNE::IModelInstanceGPU > UE::NNE::IModelGPU::CreateModelInstanceGPU ( )

pure virtual

Create a model instance for inference

The runtime have the opportunity to share the model weights among multiple IModelInstanceGPU created from an IModelGPU instance, however this is not mandatory. The caller can decide to convert the result into a shared pointer if required (e.g. if the model needs to be shared with an async task for evaluation).

Returns

A caller owned model representing the neural network instance created.

---

The documentation for this class was generated from the following file:

• Engine/Source/Runtime/NNE/Public/NNERuntimeGPU.h