Documentation for mbrl-lib¶
mbrl-lib is library to facilitate research on Model-Based Reinforcement Learning.
Getting started¶
Installation¶
mbrl-lib is a Python 3.7+ library. To install it, clone the repository,
git clone https://github.com/facebookresearch/mbrl-lib.git
then run
cd mbrl-lib
pip install -e .
If you also want the developer tools for contributing, run
pip install -e ".[dev]"
Finally, make sure your Python environment has PyTorch (>= 1.7) installed with the appropriate CUDA configuration for your system.
To test your installation, run
python -m pytest tests/core
Mujoco¶
Mujoco is a popular library for testing RL methods. Installing Mujoco is not required to use most of the components and utilities in MBRL-Lib, but if you have a working Mujoco installation (and license) and want to test MBRL-Lib on it, you please install
pip install -r requirements/mujoco.txt
and to test our mujoco-related utilities, run
python -m pytest tests/mujoco
Basic Example¶
As a starting point, check out our tutorial notebook on how to write the PETS algorithm (Chua et al., NeurIPS 2018) using our toolbox, and running it on a continuous version of the cartpole environment. Then, please take a look at our API documentation below.