All Tools
I
AgentsFreeOpen Source
IMITATION
PyTorch imitation and reward learning algorithms
MIT
ABOUT
Sparse-reward RL is slow when the only signal is a rare success, and many papers reimplement GAIL or AIRL with incompatible APIs. Imitation ships documented PyTorch algorithms that learn from expert trajectories, so teams can clone demonstrations or recover rewards without rewriting the stack.
INSTALL
pip install imitationINTEGRATION GUIDE
1. Train behavioral cloning on expert Gymnasium rollouts when reward design is hard
2. Run GAIL or AIRL to recover a reward function from demonstrations
3. Compare imitation algorithms against PPO on the same environment with shared data
4. Fine-tune a cloned policy with RL after seeding from a human or scripted expert
TAGS
pythonpytorchimitation-learninggailairlbehavioral-cloningreinforcement-learning