1. Epic GamesDeveloper
  3. ULearningAgentsTrainer

ULearningAgentsTrainer

The [ULearningAgentsTrainer](API\Plugins\LearningAgentsTraining\ULearningAgentsTrainer) is the core class for reinforcement learning training.

API > API/Plugins > API/Plugins/LearningAgentsTraining

Inheritance Hierarchy

References

   
Module LearningAgentsTraining
Header /Engine/Plugins/Experimental/LearningAgents/Source/LearningAgentsTraining/Public/LearningAgentsTrainer.h
Include #include "LearningAgentsTrainer.h"

Syntax

UCLASS&40;Abstract, BlueprintType, Blueprintable&41;  
class ULearningAgentsTrainer : public ULearningAgentsManagerComponent

Remarks

The ULearningAgentsTrainer is the core class for reinforcement learning training. It has a few responsibilities: 1) It keeps track of which agents are gathering training data. 2) It defines how those agents' rewards, completions, and resets are implemented. 3) It provides methods for orchestrating the training process.

To use this class, you need to implement the SetupRewards and SetupCompletions functions (as well as their corresponding SetRewards and SetCompletions functions), which will define the rewards and penalties the agent receives and what conditions cause an episode to end. Before you can begin training, you need to call SetupTrainer, which will initialize the underlying data structures, and you need to call AddAgent for each agent you want to gather training data from.

Constructors

  Type Name Description
Public function   Setup
Public function    

Destructors

  Type Name Description
Public function Virtual    

Functions

  Type Name Description
Public function void Used by objects derived from ULearningAgentsCompletion to add themselves to this trainer during their creation.
Public function void Used by objects derived from ULearningAgentsReward to add themselves to this trainer during their creation.
Public function UFunction BlueprintCallable, Category void

BeginTraining ( const FLearningAgentsTrainerTrainingSettings& TrainerTrainingSettings,
const FLearningAgentsTrainerGameSettings& TrainerGameSettings,
const FLearningAgentsTrainerPathSettings& TrainerPathSettings,
const FLearningAgentsCriticSettings& CriticSettings,
const bool bReinitializePolicyNetwork,
const bool bReinitializeCriticNetwork,
const bool bResetAgentsOnBegin
)

 Begins the training process with the provided settings.
Public function UFunction BlueprintCallable, Category void Stops the training process.
Public function UFunction BlueprintCallable, Category void Call this function when it is time to evaluate the completions for your agents.
Public function UFunction BlueprintCallable, Category void Call this function when it is time to evaluate the rewards for your agents.
Public function Const UFunction BlueprintPure, Category, Meta float

GetReward ( const int32 AgentId
)

 Gets the current reward for an agent according to the critic.
Public function Const UFunction BlueprintPure, Category bool Returns true if the trainer has failed to communicate with the external training process.
Public function Const UFunction BlueprintPure, Category, Meta bool

IsCompleted ( const int32 AgentId,
ELearningAgentsCompletion& OutCompletion
)

 Gets if the agent will complete the episode or not according to the given set of completions.
Public function Const UFunction BlueprintPure, Category const bool Training Process
Public function UFunction BlueprintCallable, Category void Call this function at the end of each step of your training loop.
Public function UFunction Category blueprintnativeevent void

ResetEpisodes ( const TArray< int32 >& AgentIds
)

 Resets
Public function UFunction BlueprintCallable, Category void

RunTraining ( const FLearningAgentsTrainerTrainingSettings& TrainerTrainingSettings,
const FLearningAgentsTrainerGameSettings& TrainerGameSettings,
const FLearningAgentsTrainerPathSettings& TrainerPathSettings,
const FLearningAgentsCriticSettings& CriticSettings,
const bool bReinitializePolicyNetwork,
const bool bReinitializeCriticNetwork,
const bool bResetAgentsOnBegin
)

 Convenience function that runs a basic training loop.
Public function UFunction Category blueprintnativeevent void

SetCompletions ( const TArray< int32 >& AgentIds
)

 During this event, all completions should be set for each agent.
Public function UFunction Category blueprintnativeevent void

SetRewards ( const TArray< int32 >& AgentIds
)

 During this event, all rewards/penalties should be set for each agent.
Public function UFunction Category blueprintnativeevent void Completions
Public function UFunction Category blueprintnativeevent void Rewards
Public function UFunction BlueprintCallable, Category void Initializes this object and runs the setup functions for rewards and completions.

Overridden from ULearningAgentsManagerComponent

  Type Name Description
Public function Virtual void

OnAgentsAdded ( const TArray< int32 >& AgentIds
)

 Called whenever agents are added to the parent ALearningAgentsManager object.
Public function Virtual void

OnAgentsRemoved ( const TArray< int32 >& AgentIds
)

 Called whenever agents are removed from the parent ALearningAgentsManager object.
Public function Virtual void

OnAgentsReset ( const TArray< int32 >& AgentIds
)

 Called whenever agents are reset on the parent ALearningAgentsManager object.

Overridden from UActorComponent

  Type Name Description
Public function Virtual void

EndPlay ( const EEndPlayReason::Type EndPlayReason
)

 Will automatically call EndTraining if training is still in-progress when play is ending.

See Also

ULearningAgentsInteractor to understand how observations and actions work.

Ask questions and help your peers Developer Forums

Write your own tutorials or read those from others Learning Library

On this page