unreal.MLAdapterAgent_Inference

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

Bases: MLAdapterAgent

Inference agents have a neural network that can process senses and output actuations via their Think method. You can create a blueprint of this class to easily wire-up an agent that functions entirely inside the Unreal Engine. See: UMLAdapterNoRPCManager - we do not usually need to open the RPC interface if we are exclusively using inference agents.

C++ Source:

  • Plugin: MLAdapter

  • Module: MLAdapter

  • File: MLAdapterAgent_Inference.h

Editor Properties: (see get_editor_property/set_editor_property)

  • action_duration_elapsed (bool): [Read-Only]

  • action_duration_seconds (float): [Read-Write] How long should agents wait before they can change their action

  • actuators (Array[MLAdapterActuator]): [Read-Write]

  • avatar_class (type(Class)): [Read-Write]

  • current_action_duration (float): [Read-Only] How much time has the current action been executing

  • enable_action_duration (bool): [Read-Write] If true, then agents won’t be able to make a new decision until each action duration has elapsed.

  • model_data (NNEModelData): [Read-Write] Reference to neural network asset data.

  • sensors (Array[MLAdapterSensor]): [Read-Write]