Struct FTrainerState

struct FTrainerState

A Struct representing the state of the agent given by a Reward, a vector observation and a status.

Public Functions

inline void ToProto(Schola::AgentState &OutState) const

Fill a protobuf message (Schola::AgentState) with the agent’s state.

Parameters:

OutState – [out] The protobuf message reference to fill

inline void ToProto(Schola::AgentState *OutState) const

Fill a protobuf message (Schola::AgentState) with the agent’s state.

Parameters:

OutState – [out] The protobuf message ptr to fill

inline Schola::AgentState *ToProto() const

Convert this object to a protobuf message (Schola::AgentState)

Returns:

A new protobuf message containing the contents of this object

inline bool IsDone() const

Is this agent done the current episode of training.

Returns:

true iff the agent is done it’s current episode.

inline void ToResetProto(Schola::InitialAgentState &OutState) const

Create a protobuf message (Schola::InitialAgentState) corresponding to the initial state of the agent after a reset.

Parameters:

OutState – [out] The protobuf message reference to fill

Public Members

float Reward = 0.0

The reward received by the agent in the last step.

ETrainingMsgStatus TrainingMsgStatus = ETrainingMsgStatus::NoStatus

Whether we have sent out our last message after completing an episode.

FDictPoint *Observations

The vector observation of the agent in the last step.

Not a UProperty because FDictPoint is not blueprintable

TMap<FString, FString> Info

A map of key-value pairs containing additional information about the agent from the last step.

EAgentTrainingStatus TrainingStatus = EAgentTrainingStatus::Running

The status of the agent in the last step.