API Documentation

Python API Documentation

---

Core

schola.core	Core functionality of Schola, including the main environment, spaces, and exceptions.

Extensions

schola.sb3	Support for Stable Baselines 3 environments.
schola.ray	Support for ray and rllib environments.
schola.gym	This module provides support for the Gymnasium API.

Scripts

schola.scripts	Sample scripts for running Reinforcement Learning with Schola.

Generated gRPC Code

schola.generated	Code generated by gRPC and Protobuf.

Unreal API Documentation

---

Classes

• AAbstractScholaEnvironment - An abstract class representing an environment.
• AAbstractTrainer - An abstract class representing a controller that trains an NPC using Reinforcement Learning.
• ABlueprintDynamicScholaEnvironment - A blueprintable version of the DynamicScholaEnvironment, that features trainers spawned by the environment or other objects.
• ABlueprintScholaEnvironment - A blueprintable version of the AbstractScholaEnvironment.
• ABlueprintStaticScholaEnvironment - A blueprintable version of the StaticScholaEnvironment, that features trainers spawned by the pawns they control.
• ABlueprintTrainer - A blueprint subclass of AAbstractTrainer that implements the abstract methods.
• ADynamicScholaEnvironment
• AInferenceController - A controller that implements the IInferenceAgent interface, to control a Pawn with a Brain/Policy.
• AInferencePawn - A pawn that implements the IInferenceAgent interface.
• AStaticScholaEnvironment
• CallData - A class that manages the state of a generic RPC call.
• ConstPointVisitor - A base class for objects that can operate on any const point.
• ExchangeCallData
• ExchangeRPCBackend
• ExchangeRPCWorker
• FScholaModule
• IBinaryActuatorWrapper
• IBinaryObserverWrapper
• IBlueprintBinaryActuatorWrapper
• IBlueprintBinaryObserverWrapper
• IBlueprintBoxActuatorWrapper
• IBlueprintBoxObserverWrapper
• IBlueprintDiscreteActuatorWrapper
• IBlueprintDiscreteObserverWrapper
• IBoxActuatorWrapper
• IBoxObserverWrapper
• IBrainInterface
• IComBackendInterface - An abstracted communication backend that can send string/byte messages and can either be polled for responses or do exchanges when it sends messages.
• IDiscreteActuatorWrapper
• IDiscreteObserverWrapper
• IExchangeBackendInterface - A Generic Interface for any service that can be used to exchange messages of type In and Out asynchronously.
• IGymConnector - An interface for connectors between unreal and gym.
• IInferenceAgent
• IModelInstanceInterface - Generic Interface for a model instance wrapping NNE ModelInstaces targetted at different devices.
• IModelInterface - Generic Interface for a model wrapping NNE Models targetted at different devices.
• IPollingBackendInterface - A Generic Interface for any service that can be used to receive messages of type T asynchronously via polling.
• IProducerBackendInterface - A Generic Interface for any service that can be used to send messages of type T asynchronously.
• IRuntimeInterface - Generic Interface for a runtime wrapping NNE Runtimes targetted at different devices.
• IValidatable
• PointSerializer - A class that can serialize a point to a buffer.
• PointVisitor - A base class for objects that can operate on any point.
• PollingRPCBackend
• PollingRPCWorker
• ProducerRPCBackend
• ProducerRPCWorker
• ProtobufSerializer - A class that serializes Points into the corresponding protobuf messages.
• RPCBackend
• Singleton - A singleton template that lets you make a singleton out of a type.
• UAbstractBrain - An AbstractBaseClass for subclasses representing different ways of synchronizing policy, observations and actions.
• UAbstractEnvironmentUtilityComponent - An Abstract Base class for an ActorComponent that provides utility functions for an environment.
• UAbstractGymConnector - An abstract class for connectors between Unreal Engine and gym environments.
• UAbstractInteractor - An abstract class objects that can interact with the environment, either by observing or acting.
• UAbstractNormalizer
• UAbstractObserver - An abstract class representing an observer that can collect observations about the environment state.
• UAbstractPolicy - An abstract class representing the decision making protocol of an NPC.
• UActionClipper
• UActuatorComponent - An ActorComponent for holding onto an Actuator and providing some related utility functions.
• UActuator - An actuator is a component that can be attached to an agent to make actions.
• UAgentUIDSubsystem - A subsystem that manages assigning unique numbered ids to entities.
• UAsyncBrain
• UBTTask_AgentStep - A BTTask that runs a step of inference to determine the action taken by the model.
• UBinaryActuatorWrapper
• UBinaryActuator - An actuator that does binary actions.
• UBinaryObserverWrapper - Interface for wrapping binary observations.
• UBinaryObserver - An observer that collects binary observations about the environment state.
• UBlueprintBinaryActuatorWrapper
• UBlueprintBinaryActuator - Blueprint version of binary actuator.
• UBlueprintBinaryObserverWrapper - Blueprint interface for wrapping binary observations.
• UBlueprintBinaryObserver - A blueprintable version of UBinaryObserver.
• UBlueprintBoxActuatorWrapper
• UBlueprintBoxActuator - Blueprint version of box actuator.
• UBlueprintBoxObserverWrapper - Blueprint interface for wrapping box observations.
• UBlueprintBoxObserver - A blueprintable version of UBoxObserver.
• UBlueprintDiscreteActuatorWrapper
• UBlueprintDiscreteActuator - Blueprint version of discrete actuator.
• UBlueprintDiscreteObserverWrapper - Blueprint interface for wrapping discrete observations.
• UBlueprintDiscreteObserver - Blueprintable version of UDiscreteObserver.
• UBlueprintEnvironmentUtilityComponent - A blueprint implementable version of the AbstractEnvironmentUtilityComponent.
• UBlueprintPolicy - A policy that provides hooks for implementation via Unreal Blueprint Class.
• UBoxActuatorWrapper
• UBoxActuator - An actuator that does continuous actions, and takes inputs bounded by a box space.
• UBoxObserverWrapper - Interface for wrapping box observations.
• UBoxObserver - An observer that collects box(continuous) observations about the environment state.
• UBrainInterface
• UCPUModelWrapper - Wrapper around NNE CPU Models.
• UCPURuntimeWrapper - Wrapper around NNE CPU Runtimes.
• UCameraObserver - An observer that collects 2D observations from a camera in the environment.
• UCommunicationManager - A class that manages a gRPC server running on a specified URL.
• UDebugBinaryActuator - A debug actuator that can be used for testing actuator discovery and spaces from python.
• UDebugBinaryObserver - A Debug Version of BinaryObserver that generates random observations.
• UDebugBoxActuator - A debug actuator that can be used for testing actuator discovery and spaces from python.
• UDebugBoxObserver - A Debug version of BoxObserver that generates random observations.
• UDebugDiscreteActuator - A debug actuator that can be used for testing actuator discovery and spaces from python.
• UDebugDiscreteObserver - A Debug Version of DiscreteObserver that generates random observations.
• UDiscreteActuatorWrapper
• UDiscreteActuator - An actuator that does discrete actions.
• UDiscreteBrain
• UDiscreteObserverWrapper - Interface for wrapping discrete observations.
• UDiscreteObserver - An observer that collects discrete observations about the environment state.
• UEventObserver - A Binary observer that can be triggered by an event.
• UExternalGymConnector - An abstract class for connectors that communicate with gym using futures.
• UFrameStacker
• UGPUModelWrapper - Wrapper around NNE GPU Models.
• UGPURuntimeWrapper - Wrapper around NNE GPU Runtimes.
• UGymConnector - An interface for connectors between unreal and gym.
• UHardNormalizer
• UInferenceAgent - An interface implemented by classes that represent an inference agent.
• UInferenceComponent - A component that implements the IInferenceAgent interface, to control a Pawn with a Brain/Policy.
• UInferencePolicy - Policy that uses a trained NNE model to make decisions.
• UInteractionComponent - An abstract base class for ActorComponents in Schola containing interaction logic with the environment.
• UInteractionManager
• UModelInstanceWrapper - Generic Interface for a model wrapping NNE Models targetted at different devices.
• UMovementInputActuator
• UObservationClipper
• UPositionObserver - An observer that tracks the position of an actor.
• UPythonGymConnector - A connection to an external gym API implemented in Python, using gRPC for communication.
• URayCastObserver - An observer that casts rays and collects observations about the first object hit.
• URotationActuator
• URotationObserver - An observer that tracks the rotation of an actor.
• UScholaManagerSubsystemSettings - A class to make UERL subsystem settings available in the Game tab of project settings.
• UScholaManagerSubsystem - The core subsystem that coordinates the various parts of the UERL toolkit.
• USensor - An ActorComponent for holding onto an Observer and providing some related utility functions.
• UStatLoggerComponent
• USynchronousBrain
• UTeleportActuator
• UValidatable - A class that can be validated.
• UVelocityObserver - An observer that tracks the velocity of an actor.

Structs

• FActTickFunction - A struct that represents the act portion of the tick function for the agent.
• FAction - Struct representing an action taken by an agent.
• FAgentId - Struct containing the Id of an agent.
• FBinaryPoint
• FBinarySpace - A struct representing a Binary space (e.g.
• FBoxPoint - A point in a box(continuous) space.
• FBoxSpaceDimension - A struct representing a dimension of a box(continuous) space of possible observations or actions.
• FBoxSpace - A struct representing a box(continuous) space of possible observations or actions.
• FCommunicatorSettings - A struct to hold settings relating to external communication (e.g.
• FCustomTrainingSettings - A struct to hold settings for a custom training script.
• FDictPoint - A dictionary of points.
• FDictSpace - A struct representing a dictionary of possible observations or actions.
• FDiscretePoint
• FDiscreteSpace - A struct representing a Discrete space (e.g.
• FDynamicAgentStruct
• FEnvReset - A Struct representing an update to an environment in the form of a reset.
• FEnvStep - A Struct representing an update to an environment in the form of a step.
• FEnvUpdate - A Struct representing an update to an environment in the form of either a reset or step.
• FEnvironmentDefinition - Struct containing the properties that define an environment.
• FGenericTensorBinding - A wrapper around the TensorBinding from NNE that allows us to use it in a device agnostic fashion.
• FInferencePolicyBuffer
• FInteractionDefinition - Struct containing a definition of the inputs and outputs of a policy.
• FLaunchableScript - A struct to hold the configuration of a launchable script.
• FPoint - A point is a data point that can be passed to a model for inference.
• FPolicyDecision - A struct representing a decision made by a policy.
• FRLlibAPPOSettings - A struct to hold Asynchronous Proximal Policy Optimization(APPO) settings for an RLLib training script.
• FRLlibCheckpointSettings - A struct to hold checkpoint settings for an RLlib training script.
• FRLlibIMPALASettings - A struct to hold IMPALA settings for an RLLib training script.
• FRLlibLoggingSettings - A struct to hold logging settings for an RLlib training script.
• FRLlibNetworkArchSettings - A struct to hold network architecture settings for an RLlib training script.
• FRLlibPPOSettings - A struct to hold Proximal Policy Optimization(PPO) settings for an RLLib training script.
• FRLlibResourceSettings - A struct to hold resource settings for an RLlib training script.
• FRLlibResumeSettings - A struct to hold resume settings for an RLlib training script.
• FRLlibTrainingSettings - A struct to hold all the settings for an RLlib training script.
• FSB3CheckpointSettings - A struct to hold checkpoint settings for an SB3 training script.
• FSB3LoggingSettings - A struct to hold logging settings for an SB3 training script.
• FSB3NetworkArchSettings - A struct to hold network architecture settings for an SB3 training script.
• FSB3PPOSettings - A struct to hold PPO settings for an SB3 training script.
• FSB3ResumeSettings - A struct to hold resume settings for an SB3 training script.
• FSB3SACSettings - A struct to hold SAC settings for an SB3 training script.
• FSB3TrainingSettings - A struct to hold all the settings for an SB3 training script.
• FScriptArgBuilder - A class to build arguments and flags for a CLI script.
• FScriptSettings - All settings for autolaunching a script when starting the game.
• FSharedEnvironmentDefinition - Struct containing the properties that define an environment.
• FSharedEnvironmentState - A struct representing the state of an environment given by a set of AgentStates.
• FSharedTrainingDefinition - Struct containing the properties that define a training session.
• FSpace - A class representing a space of possible observations or actions.
• FStartRequest - A struct representing a message indicating that the connector should start.
• FThinkTickFunction - A struct that represents the think portion of the tick function for the agent.
• FTrainerAgentPair
• FTrainerConfiguration - Struct containing the properties that define a trainer.
• FTrainerDefinition - Struct containing the properties that define an agent.
• FTrainerState - A Struct representing the state of the agent given by a Reward, a vector observation and a status.
• FTrainingDefinition - Struct containing the properties that define a training session.
• FTrainingSettings - Abstract class for any training settings.
• FTrainingStateUpdate - A Struct representing an update to the overall training state.
• FTrainingState - A Struct representing the state of the training session given by a set of EnvironmentStates.
• FValidationResult - Struct contraining validation result and its causer if unsuccessful.

Namespaces

• ProtobufDeserializer - A namespace containing functions to deserialize protobuf messages into Unreal Engine types.