piSTAR Lab is a modular deep reinforcement learning platform built to make AI experimentation accessible and fun.
- Web UI
- Extension System for adding new agents, environments or tasks types
- Python API, anthing you can do in the UI, you can do in Python as well
- Run agents in single and multi player environments
- Experiment tracking
- Uses Ray Project (https://ray.io/) under the hood for distributed processing
- Includes piSTAR Landia a hackable Multi Agent Envrionment
- More to come
- Cluster mode is under development and not recommended at this time
- Running remotely requires SSH tunneling. All services must be running on localhost
- Mac not tested
More detailed documentation is available at https://pistarlab.readthedocs.io
- Ubuntu suggested but also tested on Windows 10. We suggest using Docker for other setups.
- Python 3.7 or 3.8 (Conda is recommended)
- FFMPEG (Optional)
- Required for episode recordings
- Windows install instructions
- Xvfb (Optional, Ubuntu Only)
- Helps some environments run without opening a window.
- Useful when running piSTAR Lab remotely
For non-standard installations see: https://pistarlab.readthedocs.io/en/latest/installation.html
Create and Activate Conda Virtual Environment
conda create -n pistarlab python=3.7 conda activate pistarlab conda install pip
Install with pip
pip install https://github.com/pistarlab/pistarlab/archive/refs/heads/main.zip#egg=pistarlab[all]
To launch piSTAR Lab UI, run:
Open browser to: http://localhost:7777
We are still in an early phase of this release but if you are interested in contributing to piSTAR Lab, please reach out.