HumanCompatibleAI/imitation
A PyTorch library providing clean implementations of imitation and reward learning algorithms for training agents from demonstrations.
Not currently ranked — collecting fresh signals.
star history
This library provides research-quality implementations of imitation learning algorithms including Behavioral Cloning, DAgger, GAIL, and AIRL. It enables training policy models using expert demonstration data instead of hand-engineered reward functions. Built on PyTorch and integrated with Gymnasium, it offers both discrete and continuous action space support along with reward learning capabilities for inverse reinforcement learning tasks.
Frequently asked
- What is HumanCompatibleAI/imitation?
- A PyTorch library providing clean implementations of imitation and reward learning algorithms for training agents from demonstrations.
- Is imitation open source?
- Yes — HumanCompatibleAI/imitation is open source, released under the MIT license.
- What language is imitation written in?
- HumanCompatibleAI/imitation is primarily written in Python.
- How popular is imitation?
- HumanCompatibleAI/imitation has 1.8k stars on GitHub.
- Where can I find imitation?
- HumanCompatibleAI/imitation is on GitHub at https://github.com/HumanCompatibleAI/imitation.