This is the folder for the Python kit. Please make sure to read the instructions as they are important regarding how you will write a bot and submit it to the competition.
Make sure to check our Discord or the Kaggle forums for announcements if there are any breaking changes.
You will need Python 3.7 or higher and NumPy installed (which should come with the dependencies you installed for the environment)
To get started, download this folder from this repository.
Your core agent code will go into agent.py, and you can create and use more files to help you as well. You should leave main.py alone as that code enables your agent to compete against other agents locally and on Kaggle.
To quickly test run your agent, run
luxai-s2 main.py main.py --out=replay.json
This will run the agent.py code in the same folder as main.py and generate a replay file saved to replay.json.
Now that you have the code up and running, you are ready to start programming and having some fun!
If you haven't read it already, take a look at the design specifications for the competition. This will go through the rules and objectives of the competition. For a in-depth tutorial, we provide a jupyter notebook both locally and on Kaggle
All of our kits follow a common API through which you can use to access various functions and properties that will help you develop your strategy and bot. The markdown version is here: https://github.com/Lux-AI-Challenge/Lux-Design-S2/blob/main/kits/README.md, which also describes the observation and action structure/spaces.
Submissions need to be a .tar.gz bundle with main.py at the top level directory (not nested). To create a submission, create the .tar.gz with tar -czvf submission.tar.gz *. Upload this under the My Submissions tab and you should be good to go! Your submission will start with a scheduled game vs itself to ensure everything is working before being entered into the matchmaking pool against the rest of the leaderboard.
As questions come up, this will be populated with frequently asked questions regarding the Python kit.