Gym wrapper for DeepMind Lab environments

This is a gym-like wrapper for Deepmind lab environment. This repo is inspired by jkulhanek/gym-deepmindlab-env and I add some observation options which might helpful!

Usage

According to Deepmind lab's documents, we defined some task-specific wrappers:

environment id	Class names
DeepmindLabEnv-v0	DeepmindLabEnvironment
DeepmindLabNavEnv-v0	DeepmindLabMazeNavigationEnvironment

Please note that DeepmindLabMazeNavigationEnvironment is a class extends DeepmindLabEnvironment and implements abstract methods in base class.

Gym-like abstract methods:

• reset
• step
• render
• get_action_meanings

Common use:

import gym
import deepmind_lab_as_gym

env = gym.make(id='DeepmindLabNavEnv-v0', level=level)

# Use the environment
observation = env.reset()

Installation

python setup.py develop  # for development
python setup.py install  # for users

Thanks

• jkulhanek/gym-deepmindlab-env

References

• Deepmind lab: Python API
• Deepmind lab: Debug observation
• Deepmind lab: Custom observations