Expand description

Core components for reinforcement learning.


Errors in the library.

Types for recording various values obtained during training and evaluation.

A generic implementation of replay buffer.

Utilities for interaction of agents and environments.


Defines a struct that implements Shape.


All information given at every step of agent-envieronment interaction.

Gets an Agent interacts with an Env and takes samples.

Manages training loop and related objects.

Configuration of Trainer.


A set of actions of the environment.

Represents a trainable policy on an environment.

A batch of samples.

Represents an environment, typically an MDP.

Additional information to Obs and Act.

A set of observations of an environment.

A policy on an environment.

Interface of replay buffers.

Shape of observation or action.

Process Step and output an item.