GHMM: General Hidden Markov Model library: Installation

Installing GHMM

Prerequisites

Recent GCC, recent, working autoconf and automake. If you are running a recent Linux distribution you should be fine. On the MacOS we recommend using Fink to install all required packages including Python.
Libxml2
Optional: The GHMM is self-contained, but can make use of the GSL. Note that the GSL is GPL licensed and that will AFAIK force your code under the GPL. You need to pass --enable-gsl option to configure to use the GSL, in particular the nice collection of random number genetration.
For the optional Python wrappers you need the following. If building without Python wrapper use the --without-python option for configure.
- Python version 2.6.x or 2.7.x. GHMM is not yet compatible with Python 3.x releases nor with earlier releases. Note that on many Linux distributions the default installs are missing header files. You need to add the developer Python packages (which might be called python27-dev for example) and shared libs.
- SWIG, the simplified wrapper interface generator
- Optional: If you want to use the editor HMMEd you need X11 (or equivalent), Tk and Tkinter and also the Gato package. Please use the Gato out of svn. You can check it out with svn checkout svn://svn.code.sf.net/p/gato/code/trunk/Gato Gato The Gato 1.1.2 package from Sourceforge or the Python Package Index contain a bug only affecting HMMEd.

Download

Please use the version out of the SVN. There are detailed instructions at http://sourceforge.net/svn/?group_id=67094. For the most recent version you may do an anonymous checkout by calling

svn checkout svn://svn.code.sf.net/p/ghmm/code/trunk/ghmm ghmm

Installation

In the ideal case, installation is the same as for all GNU software from source: In the ghmm directory issue the following commands.


> autogen.sh
> ./configure
> make 
> make install

This will install the GHMM into whatever path prefix is default for autoconf on your system. Note that the same prefix is used for installing the Python library, the path for the Python install is not determined by the install location of the first python found on your PATH. That is, if for example /usr/local is the default for autoconf, the ghmm.py will end up in /usr/local/lib/pythonX.Y/site-packages. This will cause an import error if your python is not installed in /usr/local or /usr/local/lib/pythonX.Y/site-packages is not added to your PYTHONPATH environment variable.

If your Python is in a non-standard location for autoconf, you probably want to install into the appropriate site-packages. That is, if the Python interpreter invoked by typing python in your shell is /var/bin/python than most likely its library will be /var/lib/pythonX.Y and all site-specific installs should end up in /var/lib/pythonX.Y/site-packages/ by default.

There is a simple work-around as follows.


> autogen.sh
> ./configure --without-python
> make 
> make install
> cd ghmmwrapper
> python setup.py build
> python setup.py install

Sometimes a more fine-grained control is needed. Then you can specify the prefixes used for installation and the python interpreter to use explicitely.


> autogen.sh
> ./configure --without-python --prefix=/some/path
> make 
> make install
> cd ghmmwrapper
> /some/otherpath/python setup.py build
> /some/otherpath/python setup.py install --prefix=/yetanotherpath/

This might require setting your PYTHONPATH or LD_LIBRARY_PATH environment variable appropriately.

The graphical editor can be installed after installing Gato (see above) just like the Python bindings in ghmmwrapper


> cd ../HMMEd
> /some/otherpath/python setup.py build
> /some/otherpath/python setup.py install --prefix=/yetanotherpath/

Ubuntu installation walkthrough

First make sure you have the prerequisites


sudo apt-get update 
sudo apt-get install build-essential automake autoconf libtool
sudo apt-get install python-dev
sudo apt-get install libxml++2.6-dev
sudo apt-get install swig

Then get the tarball and unpack or checkout from svn and run


cd ghmm
sh autogen.sh
sudo ./configure
sudo make
sudo make install
sudo ldconfig

OS-specific tips

Mac OS X: 10.6 ships with a broken libtool which breaks the installation (and it also ships with Python 2.5, so you need an update for that). James Howard posted a solution on the mailing list: [Ghmm-list] Compiling in OS X 10.6
Windows: There is no Windows support. Installing Cygwin should work. Check the mailing list; it seems that some people had success.
Linux: various distributions are creative about install locations and the scope of what actually is installed. You want the developer/header whatever-they-are-called packages in addition to that. Note that specific distributions also omit part of the Python standard library (e.g., Ubuntu omits the Python profiler for licensing reasons).
C++: We used to have C++ bindings emulating the semantics of the Python interface, but we lost interest after it became clear that Python plus C is nicer and as fast.

Troubleshooting

autogen.sh

configure

libtool: Version mismatch error.

cautoreconf --force --install

configure; make; make install ...

There are more options to configure, check configure --help. Note that --disable-obsolete is currently broken.
There is anecdotal evidence that adding search paths to ghmm/ghmmwrapper/setup.py is the easiest way to get around install problems in really creative installations.
If you don't have root access, installing somewhere in your home directory by specifying the --prefix arguments seems to be the path of least resistence. Then you need to set PYTHONPATH, LD_LIBRARY_PATH and possibly PATH accordingly in your .bashrc.