Snake Lab
Snake Lab is built around Patrick Loeber’s Train an AI to Play Snake Tutorial. The reinforcement learning tutorial demonstrates how to build a simple AI that learns to play Snake by training over hundreds or thousands of games.
At the heart of the AI is its neural network and training configuration. Changes to parameters such as model size, learning rate, discount factor, and reward values can have a significant effect on the performance the AI ultimately achieves.
Snake Lab accepts a simulation configuration message over the network. The message describes the neural network and training configuration, along with instructions for running the simulation. Snake Lab validates the request, executes the simulation, and stores the results in a database where clients can retrieve the data.
This project was created to support the Ax3l Project, which uses a Large Language Model (LLM) to optimize the configuration and maximize the high score.