Metadata-Version: 2.4
Name: mjx-safety-gym
Version: 0.1.0
Summary: MJX Safety Gym Environments
Author-email: Max van der Hart <your.email@example.com>
License-Expression: MIT
Project-URL: Homepage, https://github.com/yourusername/mjx-safety-gym
Requires-Python: >=3.11
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: absl-py<3.0.0,>=2.2.2
Requires-Dist: etils==1.13.0
Requires-Dist: flax==0.10.6
Requires-Dist: jax==0.6.0
Requires-Dist: ml_collections==1.1.0
Requires-Dist: mujoco==3.3.2
Requires-Dist: mujoco_mjx==3.3.2
Requires-Dist: numpy==2.3.2
Requires-Dist: pynput==1.8.1
Requires-Dist: setuptools==65.5.0
Dynamic: license-file

# mjx-safety-gym

## Setup
Create a Python venv with version 3.11 (or above). 

Then, 
```pip install -r requirements.txt```
After whcih 
```pip install -e .``` 
To install the module 

If you want to use vision-based observations, see the Madrona section first. 

## How to Run 
Open-source MJX implementation of OpenAI Safety Gym for accelerated safe reinforcement learning.

To run the interactive viewer on MacOS M1, run
```sudo mjpython scripts/interactive.py ```

Note that we need sudo privileges to get access to keypress information. 


## Madrona
To use vision-based observations, run 
```
chmod +x vision_setup.bash
./vision_setup.bash
```
This requires a linux machine with an NVidia GPU and can take up to 5 minutes to install.

