Welcome to Tianshou!#

Tianshou (天授) is a reinforcement learning platform based on pure PyTorch. Unlike existing reinforcement learning libraries, which are mainly based on TensorFlow, have many nested classes, unfriendly API, or slow-speed, Tianshou provides a fast-speed framework and pythonic API for building the deep reinforcement learning agent. The supported interface algorithms include:

• DQN Deep Q-Network

• DQN Double DQN

• DQN Dueling DQN

• BDQN Branching DQN

• C51 Categorical DQN

• RainbowDQN Rainbow DQN

• QRDQN Quantile Regression DQN

• IQN Implicit Quantile Network

• FQF Fully-parameterized Quantile Function

• Reinforce Reinforce/Vanilla Policy Gradients

• NPG Natural Policy Gradient

• A2C Advantage Actor-Critic

• TRPO Trust Region Policy Optimization

• PPO Proximal Policy Optimization

• DDPG Deep Deterministic Policy Gradient

• TD3 Twin Delayed DDPG

• SAC Soft Actor-Critic

• REDQ Randomized Ensembled Double Q-Learning

• DiscreteSAC Discrete Soft Actor-Critic

• ImitationPolicy Imitation Learning

• BCQ Batch-Constrained deep Q-Learning

• CQL Conservative Q-Learning

• TD3BC Twin Delayed DDPG with Behavior Cloning

• DiscreteCQL Discrete Conservative Q-Learning

• DiscreteBCQ Discrete Batch-Constrained deep Q-Learning

• DiscreteCRR Critic Regularized Regression

• GAIL Generative Adversarial Imitation Learning

• PSRLPolicy Posterior Sampling Reinforcement Learning

• ICMOffPolicyWrapper, ICMOnPolicyWrapper Intrinsic Curiosity Module

• PrioritizedReplayBuffer Prioritized Experience Replay

• compute_episodic_return() Generalized Advantage Estimator

• HERReplayBuffer Hindsight Experience Replay

$ pip install tianshou

$ pip install git+https://github.com/thu-ml/tianshou.git@master --upgrade

import tianshou
print(tianshou.__version__)