jonzamora / dmc2gymnasium

Farama Gymnasium API Wrapper for the DeepMind Control Suite and DeepMind Robot Manipulation Tasks

dmc2gymnasium

This is a lightweight wrapper around the DeepMind Control Suite and DeepMind Robot Manipulation Tasks, and provides the standard Farama Gymnasium API interface to users.

Farama Gymnasium is the continuation of OpenAI Gym, and this repository will provide users with a simple, up-to-date, and easy-to-install package for their DM Control Suite experiments.

For a complete list of tasks, see the dm_control paper here and check out Part II of the Paper.

Installation

git clone https://github.com/jonzamora/dmc2gymnasium
cd dmc2gymnasium
pip install -e .

Usage

A complete usage of the DMCWrapper is as follows:

from dmc2gymnasium import DMCWrapper

env = DMCWrapper(domain_name="manipulation", task_name="place_brick_features")
observation, info = env.reset(seed=42)

for step in range(1000):
    action = env.action_space.sample()
    observation, reward, terminated, truncated, info = env.step(action)

    if terminated or truncated:
        observation, info = env.reset(seed=42)

env.close()

If you'd like to test this out, simply run the following:

python test.py

Output will look like this in the video directory:

demo.mp4

Notes

I've mostly designed this for my own experiments, but feel free to use this if it applies to your work!

If you have a feature request, or encounter any problems, please make an Issue here and i'm happy to discuss further.

References

This work is largely inspired by the following repos:

• imgeorgiev/dmc2gymnasium [Link]
• denisyarats/dmc2gym [Link]