imitation 1.0.0


Implementation of modern reward and imitation learning algorithms.

Implementation of modern reward and imitation learning algorithms.

Stars: 1127, Watchers: 1127, Forks: 218, Open Issues: 78

The HumanCompatibleAI/imitation repo was created 5 years ago and the last code push was 3 weeks ago.
The project is very popular with an impressive 1127 github stars!

How to Install imitation

You can install imitation using pip

pip install imitation

or add it to a project with poetry

poetry add imitation

Package Details

Center for Human-Compatible AI and Google
GitHub Repo:


No  imitation  pypi packages just yet.


A list of common imitation errors.

Code Examples

Here are some imitation code examples and snippets.

GitHub Issues

The imitation package has 78 open issues on GitHub

  • Gymnasium compatibility maintaining python3.8
  • Don't train experts in the tutorials, download from HF instead
  • Ensure all tutorials work as expected
  • prob_true_act > 1 with PPO and BC?
  • more cli examples
  • Refactor trajectory handling
  • The documentation and code in the preference comparison tutorials has become out of sync
  • Approximate a finite horizon environment?
  • Add tutorials for the common use cases of the CLI
  • Implementation of the SQIL algorithm
  • Implement Soft Q imitation learning (SQIL)
  • comptibilte between gym and gymnasium
  • Gymnasium Compatibility
  • gymnasium compatibility of imitation
  • Indexing a TrajectoryDatasetSequence instance when observations are images is extremely slow.

See more issues on GitHub