unreal.LearningNeuralNetworkData

class unreal.LearningNeuralNetworkData(outer: Object | None = None, name: Name | str = 'None')

Bases: Object

Neural Network Data Object

This is the UObject which contains the actual data used by a Neural Network. It stores the raw FileData used to construct the network using NNE, as well as the input and output sizes and a compatibility hash that can be used to quickly check if two networks may be compatible in terms of inputs and outputs.

Internally this also stores the various things required to map between NNE style inference and the style of inference used in Learning via FNeuralNetwork and FNeuralNetworkInference.

C++ Source:

  • Plugin: LearningAgents

  • Module: Learning

  • File: LearningNeuralNetwork.h

Editor Properties: (see get_editor_property/set_editor_property)

  • compatibility_hash (int32): [Read-Only] Compatibility hash used for testing if inputs to one network are compatible with another

  • input_size (int32): [Read-Only] Size of the inputs expected by this network

  • output_size (int32): [Read-Only] Size of the output produced by this network