Skip to content

Latest commit

 

History

History
59 lines (44 loc) · 2.36 KB

readme.md

File metadata and controls

59 lines (44 loc) · 2.36 KB

AI Snake

This project created using python. Snake game was created using pygame library, NN using pyTorch.
You can easily test it just by completing following steps:

Test project:

        * git clone https://github.com/kinfi4/AISnake.git
        * cd Snake
        * python3 -m venv env && source env/bin/activate
        * pip install -r requirements.txt
        * python test.py model/model.trained.pth

This project has two main scripts: train.py and test.py

train.py:

Trains the model, visualize the snake and plots its scores and mean of its scores.  
This script takes two optional arguments: 
    -f (--filename) <path to file> - file where script will score trained NN,
        default: ./model/model.pth
    -s (--short_form) - if is it set, scripts will not show you plots

Examples:
    python train.py ./model/model.trained.pth -- train model and save it the specific path 
    python train.py ./model/model.trained.pth -s -- train model without plotting

test.py

Tests the model, just visualize the snake playing.
This script takes one required argument and one optional:
    model_file - required positional arguments - path to the trained model
    -s (--speed) - optional argument, using it you can specify speed of the snake

Examples:
    python test.py ./model/model.trained.pth
    python test.py ./model/model.trained.pth -s 100 -- play game with speed 100

Architecture of NN

NN has three layers: input, hidden and output with (11, 256, 3) neurons respectively.
Input layer gets game state, and build prediction based on it.
Input later consists of 11 numbers:

  • Does the snake has danger straight, right or left from her (3 numbers)
  • Direction of snake movement, right, down, left or up (4 numbers)
  • Place of food, left, right, down and up from snake (4 numbers)

Screenshots:

@kinfi4