All Tools
A
AgentsFreeOpen Source
ARCADE LEARNING ENVIRONMENT
Atari 2600 environments for visual reinforcement learning
GPL-2.0
ABOUT
Visual RL needs a shared Atari benchmark, but raw Stella emulation is painful to wire into modern trainers. ALE gives RAM and pixel observations with a Gymnasium API so papers can compare DQN-family agents on the same 2600 games without each lab packaging the emulator.
INSTALL
pip install ale-pyINTEGRATION GUIDE
1. Train Rainbow DQN or PPO on Breakout, Pong, or Montezuma's Revenge via ale-py
2. Benchmark sample-efficient visual RL algorithms on the standard ALE suite
3. Use RAM observations when pixel-based training is too slow for a prototype
4. Plug Atari into Stable Baselines3 or Tianshou through the Gymnasium wrapper
TAGS
pythonatarireinforcement-learninggymnasiumfaramapixels