A2C\A2CAgent, A2C\ActorCritic — A2CDDPG\DDPGAgent, DDPG\Actor, DDPG\Critic — DDPGDDPG\OrnsteinUhlenbeckNoise — DDPG explorationDQN\DQNAgent, DQN\QNetwork — DQNPPO\PPOAgent, PPO\ActorCritic — PPOQLearning\QLearningAgent — Q-learningReinforce\ReinforceAgent, Reinforce\PolicyNetwork — REINFORCESAC\SACGSDEAgent, SAC\GSDEActor, SAC\Critic, SAC\QNetwork — SAC+gSDESarsa\TrueOnlineSarsaLambdaAgent, Sarsa\TileCoder — True Online Sarsa(λ)ReplayBuffer\ReplayBuffer — ReplayBufferReplayBuffer\RolloutBuffer — RolloutBufferEach agent namespace contains a Runner. See runner overview,
on-policy, and
off-policy.
Env\CartPole\DeviceWrapperEnv\MountainCar\DeviceWrapperEnv\ContinuousMountainCar\DeviceWrapperEnv\Pendulum\DeviceWrapperEnv\Maze\DeviceWrapperSee Device wrappers.
Util\ActionMask — Action masksUtil\GradientClipping — Gradient clippingUtil\TensorValidation — Tensor validationUtil\ProgressBar — Progress reporting