Skip to content

Latest commit

 

History

History
40 lines (34 loc) · 1.22 KB

README.md

File metadata and controls

40 lines (34 loc) · 1.22 KB

gym_snake

An implementation of the game 'Snake' for the OpenAI Gym framework. This is done to make it easily connect to stable-Baselines.

The environment

The Snake environment can be installed by running 'pip install .' after downloading the setup.py and gym_snake folder. You can then create the game from Gym using

import gym

env = gym.make('gym_snake:snake-v0')

or using

from gym_snake import SnakeEnv

env = SnakeEnv(**kwargs)

where the arguments are

shape : (int, int)
 This is the size of the grid

length : int (optional)
 This is the starting length of the snake (defaults to 3).

seed : int (optional)
 The seed is provided for so that runs can be reproduced.

Getting started

The script snake_trainer.py is provided as an example to both train and watch the snake after it has been trained. The snake is 'trained' using DeepQ learning. The snake cannot be watched while training, but you can tell it to train using python snake_trainer.py and then tell it to play using python snake_trainer.py --play. Training (can be) computationally intensive.