🛢️

BARL: Base Agents for Reinforcement Learning

This codebase provides implementations of many RL algorithms with the goal of being flexible for new algorithm ideas and fast experimentation.

TODO:

• Architecture kwargs (rename to model?)
• reduce branching in preprocess obs, maybe cache which preprocess for each env (map of env to function)
• add more tests
• dueling architecture
• hyperparameters
• random seeds
• double check Polyak averaging tau, timing
• Clean up eval thread

Algorithms:

• DQN
• SQL (discrete actions)
• SAC
• Rainbow DQN
• TD3
• PPO

Architectures:

• MLP
• CNN
• LSTM
• Dueling architectures