These tutorials provide an introduction to using Tianshou with PettingZoo.
Basic API Usage: View a game between random agents
Training Agents: Train a DQN agent
CLI and Logging: Full training script with CLI and logging
Tianshou is a lightweight reinforcement learning platform providing fast-speed, modularized framework and pythonic API for building the deep reinforcement learning agent with the least number of lines of code. It uses pure PyTorch and is written in only ~4000 lines of code.
It boasts a large number of algorithms and high quality software engineering standards: thorough testing, type hints, and comprehensive documentation.