1. Epic GamesDeveloper
  3. FLearningAgentsTrainerTrainingSettings

FLearningAgentsTrainerTrainingSettings

The configurable settings for the training process.

API > API/Plugins > API/Plugins/LearningAgentsTraining

References

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

Syntax

USTRUCT&40;BlueprintType, Category&61;"LearningAgents"&41;  
struct FLearningAgentsTrainerTrainingSettings

Remarks

The configurable settings for the training process.

Variables

  Type Name Description
Public variable UProperty Category, EditAnywhere, Meta float ActionRegularizationWeight Weight used to regularize actions.
Public variable UProperty Category, EditAnywhere bool bAdvantageNormalization When true, advantages are normalized.
Public variable UProperty Category, EditAnywhere, Meta int32 BatchSize Batch size to use for training.
Public variable UProperty Category, EditAnywhere bool bClipAdvantages When true, very large or small advantages will be clipped.
Public variable UProperty Category, EditAnywhere bool bUseTensorboard If true, TensorBoard logs will be emitted to the intermediate directory.
Public variable UProperty Category, EditAnywhere ELearningAgentsTrainerDevice Device The device to train on.
Public variable UProperty Category, EditAnywhere, Meta float DiscountFactor The discount factor to use during training.
Public variable UProperty Category, EditAnywhere, Meta float EntropyWeight Weighting used for the entropy bonus.
Public variable UProperty Category, EditAnywhere, Meta float EpsilonClip Clipping ratio to apply to policy updates.
Public variable UProperty Category, EditAnywhere, Meta float GaeLambda This is used in the Generalized Advantage Estimation as what is essentially an exponential smoothing/decay.
Public variable UProperty Category, EditAnywhere, Meta float InitialActionScale The initial scaling for the weights of the output layer of the neural network.
Public variable UProperty Category, EditAnywhere, Meta float LearningRateCritic Learning rate of the critic network.
Public variable UProperty Category, EditAnywhere, Meta float LearningRateDecay Ratio by which to decay the learning rate every 1000 iterations.
Public variable UProperty Category, EditAnywhere, Meta float LearningRatePolicy Learning rate of the policy network. Typical values are between 0.001 and 0.0001.
Public variable UProperty Category, EditAnywhere, Meta int32 NumberOfIterations The number of iterations to run before ending training.
Public variable UProperty Category, EditAnywhere, Meta int32 NumberOfStepsToTrimAtEndOfEpisode The number of steps to trim from the end of the episode.
Public variable UProperty Category, EditAnywhere, Meta int32 NumberOfStepsToTrimAtStartOfEpisode The number of steps to trim from the start of the episode, e.g. can be useful if some things are still getting setup at the start of the episode and you don't want them used for training.
Public variable UProperty Category, EditAnywhere, Meta int32 RandomSeed The seed used for any random sampling the trainer will perform, e.g. for weight initialization.
Public variable UProperty Category, EditAnywhere, Meta float WeightDecay Amount of weight decay to apply to the network.

Ask questions and help your peers Developer Forums

Write your own tutorials or read those from others Learning Library

On this page