amle-py 1.1.5

A package that can be used to make an AI learn from Amstrad CPC games.

Amstrad Learning Environment

The Amstrad Learning Environment is a Python interface that is meant to be used with AIs and OpenAi Gym. The Python library is available on Linux and Windows.

Installation

Unix

First you’ll need to install Python3 (version >= 3.6). To install Python :

$ sudo apt-get install python3

To check the version :

$ python3 --version

Then, you’ll need to install pip, so you can install the amle-py package :

$ curl https://bootstrap.pypa.io/get-pip.py | python

With pip, you’ll need to install two dependencies :

$ pip install --upgrade setuptools
$ pip install --upgrade numpy

Finally you can install amle-py :

$ pip install --upgrade amle-py

Windows

Install Python by going on their website. Make sure to add Python to your path and install pip during the installation.

How to use

Compile and use the sources (Linux only)

To compile the sources you will need to install a few dependencies :

• SDL 1.2

• libpng

• libzip

• cppunit

Once you have all these dependencies, you can compile the project. You may also want to generate the documentation of the project, this is done in the doc/ folder. Make sure to have doxygen installed :

$ sudo apt-get install doxygen

And then :

$ doxygen doxygen.config

Use the Python library

First you will need to create a new Python file in which you import the library :

import amle_py

Then you will need to create a new amle object :

amle = amle_py.AmLEInterface()

And you will need to load a game :

amle.loadSnapshot("Arkanoid", "snap/arka.sna")

Note that the first string has to a name the AmLE can understand. If you have a doubt you can get the list of all possible strings with :

games = amle.getSupportedGames()

Moreover, the second argument is a path to YOUR .sna file. You have to import one from the internet and generate one yourself with an emulator. Also, it doesn’t have to be in a snap/ folder, this is just cleaner.

Then you may want to run the game :

while not(amle.gameOver()):
    amle.step()

This doesn’t do anything interesting though, you can also interact with the game. To do so, before the loop you can do :

nbLegalActions = amle.getNbLegalActions()
legalActions = np.empty(nbLegalActions, dtype=np.int32)
amle.getLegalActions(legalActions)
legalActions = legalActions.tolist()

And then in the loop :

import amle_py
import random
# The previous code discussed
while not(amle.gameOver()):
    amle.act(random.choice(legalActions))
    amle.step()

Finally, you may want to generate the documentation for the library. To do so go in the amle_py folder and run :

$ pydoc -w amle_python_interface.py