Algorithms

d3rlpy provides state-of-the-art offline deep reinforcement learning algorithms as well as online algorithms for the base implementations.

Continuous control algorithms

d3rlpy.algos.BC	Behavior Cloning algorithm.
d3rlpy.algos.DDPG	Deep Deterministic Policy Gradients algorithm.
d3rlpy.algos.TD3	Twin Delayed Deep Deterministic Policy Gradients algorithm.
d3rlpy.algos.SAC	Soft Actor-Critic algorithm.
d3rlpy.algos.BCQ	Batch-Constrained Q-learning algorithm.
d3rlpy.algos.BEAR	Bootstrapping Error Accumulation Reduction algorithm.
d3rlpy.algos.CRR	Critic Reguralized Regression algorithm.
d3rlpy.algos.CQL	Conservative Q-Learning algorithm.
d3rlpy.algos.AWR	Advantage-Weighted Regression algorithm.
d3rlpy.algos.AWAC	Advantage Weighted Actor-Critic algorithm.
d3rlpy.algos.PLAS	Policy in Latent Action Space algorithm.
d3rlpy.algos.PLASWithPerturbation	Policy in Latent Action Space algorithm with perturbation layer.
d3rlpy.algos.TD3PlusBC	TD3+BC algorithm.
d3rlpy.algos.MOPO	Model-based Offline Policy Optimization.
d3rlpy.algos.COMBO	Conservative Offline Model-Based Optimization.
d3rlpy.algos.RandomPolicy	Random Policy for continuous control algorithm.

Discrete control algorithms

d3rlpy.algos.DiscreteBC	Behavior Cloning algorithm for discrete control.
d3rlpy.algos.DQN	Deep Q-Network algorithm.
d3rlpy.algos.DoubleDQN	Double Deep Q-Network algorithm.
d3rlpy.algos.DiscreteSAC	Soft Actor-Critic algorithm for discrete action-space.
d3rlpy.algos.DiscreteBCQ	Discrete version of Batch-Constrained Q-learning algorithm.
d3rlpy.algos.DiscreteCQL	Discrete version of Conservative Q-Learning algorithm.
d3rlpy.algos.DiscreteAWR	Discrete veriosn of Advantage-Weighted Regression algorithm.
d3rlpy.algos.DiscreteRandomPolicy	Random Policy for discrete control algorithm.