All Tools
P
AgentsFreeOpen Source
PUFFERLIB
High-performance vectorized reinforcement learning
MIT
ABOUT
Many RL environments cannot keep a GPU busy because stepping is slow and poorly batched. PufferLib vectorizes env execution and training so researchers can run millions of steps per second on CPU-heavy games and custom sims without writing a one-off collector.
INSTALL
pip install pufferlibINTEGRATION GUIDE
1. Train a fast PPO agent on NetHack, Pokemon, or another PufferLib-wrapped env
2. Vectorize a custom Gymnasium environment to raise sample throughput
3. Benchmark environment wrappers against EnvPool or Sample Factory collectors
4. Prototype a game-playing agent where simulation speed is the bottleneck
TAGS
pythonreinforcement-learningvectorizedgymnasiumhigh-throughputpuffer