For installing of the most recent version of the MMLF, please download the tar.gz file from https://sourceforge.net/projects/mmlf/. With this file, the MMLF can either be installed locally or globally (see below).
The MMLF has been tested on Python v2.6 and v2.7 and requires that the python packages numpy (tested on version 1.3.0), scipy (v0.7.0), matplotlib (v1.0), and pyyaml (v3.0.9) are installed. If you want to use the graphical user interface, the PyQt4 (v4.7.2) package is required additional. Specific configurations of agents may require additional packages (e.g., scikits.ann and lwpr); however, most of the MMLF should be usable when the core dependencies are fulfilled. The easiest way to install the dependencies depends on your operating system. Under Windows you may use the python(x,y) distribution. Under Mac OSX, you may use MacPorts, and under Linux, your packaging tool will most likely ship these packages out-of-the-box.
This kind of installation is most useful for developers that don’t have access to the MMLF revision control system. Installing MMLF locally means to simply extract the tar.gz archive. The resulting directory consists of the mmlf package, this documentation, the configuration directory, and two python-scripts: run_mmlf (for running MMLF in a non-graphical way) and mmlf_gui (for starting a GUI which allows to configure and run the MMLF). The additional script setup.py is required for installing the MMLF globally. Running the MMLF via the run_mlf and mmlf_gui scripts is possible without requiring any further adjustments (see Quick Start Tutorial). If the MMLF should be made available as a module to other python code, please add the directory into which you extracted the MMLF to your pythonpath. It will then be possible to use import mmlf in a python script or interactive interpreter session. Any changes made in the code in the mmlf directory are directly active (changes to config files should be made in the $HOME/.mmlf directory however – see the note below).
This kind of installation is most useful for people who don’t want to modify or extend the MMLF, but only wish to use it. A global installation allows the MMLF to be installed in a way such that all users of a system can use it. In order to install globally, extract the tar.gz archive and run python setup.py install. This command may require superuser privileges, and uses the Python Distribution Utilities (distutils) to install the MMLF into the site-packages/dist-packages directory of your python distribution. You may have to install setuptools before. After installing, you can run the mmlf via the run_mmlf (for running MMLF in a non-graphical way) and mmlf_gui (for starting a GUI which allows to configure and run the MMLF) scripts. See Quick Start Tutorial for more details. Furthermore, you can now also import and use the mmlf module in your own python scripts by import mmlf (see MMLF python package interface).
Under Windows, you have to call python mmlf_gui in the Scripts subdirectory of your python directory (e.g. C:\Python26\Scripts)
Performing a global installation copies MMLF config files into /etc/mmlf (unix-like OSes) or $APPDATA/mmlf (Windows). However, the data in these directories only defines some default world setups, and is not meant for modification. If you wish to modify a world’s setup, modify the configuration files in the $HOME/.mmlf directory that is created after the first start of the MMLF. For more information on that, see the Quick Start Tutorial