DeepMind Hard Eight Tasks is a set of 8 diverse machine-learning tasks that require exploration in partially observable environments to solve.

These tasks are provided through pre-packaged Docker containers.

This package consists of support code to run these Docker containers. You interact with the task environment via a dm_env Python interface.

Please see the documentation for more detailed information on the available tasks, actions and observations.

The Hard Eight tasks are intended to be run on Linux and are not officially supported on Mac and Windows. However, they can in principle be run on any platform. In particular, on Windows, you may need to run the Python code from within WSL.

dm_hard_eight requires Docker, Python 3.6.1 or later and a x86-64 CPU with SSE4.2 support. We do not attempt to maintain a working version for Python 2.

Note: We recommend using Python virtual environment to mitigate conflicts with your system's Python environment.

Download and install Docker:

• For Linux, install Docker-CE
• Install Docker Desktop for OSX or Windows.

You can install dm_hard_eight by cloning a local copy of our GitHub repository:

$ git clone https://github.com/deepmind/dm_hard_eight.git
$ pip install ./dm_hard_eight

To also install the dependencies for the examples/, install with:

$ pip install ./dm_hard_eight[examples]

Once dm_hard_eight is installed, to instantiate a dm_env instance run the following:

import dm_hard_eight

settings = dm_hard_eight.EnvironmentSettings(seed=123,
    level_name='ball_room_navigation_cubes')
env = dm_hard_eight.load_from_docker(settings)

If you use dm_hard_eight in your work, please cite the accompanying paper:

@article{paine2019making,
  title={Making Efficient Use of Demonstrations to Solve Hard Exploration Problems},
  author={Tom Le Paine and
          Caglar Gulcehre and
          Bobak Shahriari and
          Misha Denil and
          Matt Hoffman and
          Hubert Soyer and
          Richard Tanburn and
          Steven Kapturowski and
          Neil Rabinowitz and
          Duncan Williams and
          Gabriel Barth-Maron and
          Ziyu Wang and
          Nando de Freitas and
          Worlds Team}
  journal={arXiv preprint arXiv:1909.01387},
  year={2019}
}

This is not an officially supported Google product.