A learning environment written in C++ and Lua for the creation of grid worlds.
DeepMind Lab2D is a system for the creation of 2D environments for machine
learning. The main goals of the system are ease of use and performance: The
environments are "grid worlds", which are defined with a combination of simple
text-based maps for the layout of the world, and Lua code for its behaviour.
Machine learning agents interact with these environments through one of two
APIs, the Python dm_env API or a custom
C API (which is also used by DeepMind Lab).
Multiple agents are supported.
If you use DeepMind Lab2D in your research and would like to cite it, we suggest you cite the accompanying whitepaper.
We provide an example "random" agent in python/random_agent, which performs
random actions. This can be used as a base for creating your own agents, and as
a simple tool to preview an environment.
bazel run -c opt dmlab2d/random_agent -- --level_name=clean_upDeepMind Lab2D currently ships as source code only. It depends on a few external software libraries, which we ship in several different ways:
-
The
dm_env,eigen,luajit,lua5.1,lua5.2,luajit,pngandzliblibraries are referenced as external Bazel sources, and Bazel BUILD files are provided. The dependent code itself should be fairly portable, but the BUILD rules we ship are specific to Linux on x86. To build on a different platform you will most likely have to edit those BUILD files. -
A "generic reinforcement learning API" is included in
//third_party/rl_api. -
Several additional libraries are required but are not shipped in any form; they must be present on your system:
Python 3.6or above withNumPy,PyGame, andpackaging.
The build rules are using a few compiler settings that are specific to GCC/Clang. If some flags are not recognized by your compiler (typically those would be specific warning suppressions), you may have to edit those flags.
This is not an official Google product.