reference/objectives_common

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

Common Components

Base classes and common utilities for all loss modules.

LossModule(*args, **kwargs)	A parent class for RL losses.
add_random_module(module)	Adds a random module to the list of modules that will be detected by vmap_randomness() as random.

Value Estimators

ValueEstimatorBase(*args, **kwargs)	An abstract parent class for value function modules.
TD0Estimator(*args, **kwargs)	Temporal Difference (TD(0)) estimate of advantage function.
TD1Estimator(*args, **kwargs)	\(\infty\)-Temporal Difference (TD(1)) estimate of advantage function.
TDLambdaEstimator(*args, **kwargs)	TD(\(\lambda\)) estimate of advantage function.
GAE(*args, **kwargs)	A class wrapper around the generalized advantage estimate functional.
VTrace(*args, **kwargs)	A class wrapper around V-Trace estimate functional.
MultiAgentGAE(*args, **kwargs)	Multi-agent Generalized Advantage Estimator.

ValueEstimators(value[, names, module, ...])

Value function enumerator for custom-built estimators.