| Main developer: | Thejs Brinckmann <brinckmann@physik.rwth-aachen.de> |
|---|---|
| Author: | Benjamin Audren <benjamin.audren@epfl.ch> |
| License: | MIT |
The code is under the MIT license. As an additional clause, when using the code
in a scientific publication you are also required to cite the v3.0 release paper
MontePython 3: boosted MCMC sampler and other features and the original release
paper Conservative Constraints on Early Cosmology (see the tail of this document
for the bibtex entries).
If you are searching for specific examples of a work session, please refer to the pdf documentation. See also the MontePython 3 paper for details on the code, as well as a summary of features as of v3.0. This includes an overview of all likelihoods currently implemented, with some details on those likelihoods, such as adatasets, last updated, type and relevant papers to cite when using the likelihood.
- You need the python program version 2.7 or above, but less than 3.0. Note that lower versions of python will work, down to 2.6 (tested), if you add manually two extra packages ( ordereddict and argparse).
- Your python of choice must have numpy (version >= 1.4.1) and cython. The later is used to wrap CLASS in python.
- [optional] If you want to use fully the plotting capabilities of Monte Python, you also need the scipy, with interpolate, and matplotlib modules.
- [optional] You can now use Multi Nest and the CosmoHammer with Monte Python, though you need to install them. Please refer to the documentation.
Move the .tar.bz2 file to the place of your convenience, untar its content
$ bunzip2 montepython-vx.y.tar.bz2 $ tar -xvf montepython-vx.y.tar
This will create a directory named montepython into your current directory. You can add the following line to your .bashrc file:
export PATH=/path/to/MontePython/montepython/:$PATH
to be able to call the program from anywhere.
You will need to adapt only two files to your local configuration. The first is the main file of the code montepython/MontePython.py, and it will be the only time you will have to edit it, and it is simply to accommodate different possible configurations of your computer.
Its first line reads
#!/usr/bin/python
This should be changed to wherever is your preferred python distribution installed. For standard distribution, this should already be working. Now, you should be able to execute directly the file, i.e. instead of calling:
The second file to modify is located in the root directory of Monte Python : default.conf. This file will be read (and stored) whenever you execute the program, and will search for your cosmological code path, your data path, and your wmap wrapper path. You can alternatively create a second one, my.conf, containing your setup, and then run the code providing this file (with the flag --conf)
Go to your class directory, and do make clean, then make. This builds the libclass.a, needed for the next step. From there,
$ cd python/ $ python setup.py build $ python setup.py install --user
This will compile the file classy.pyx, which is the python wrapper for CLASS, into a library, classy.so, located in the build/ subdirectory. This is the library called in Monte Python afterwards.
If this step fails, check that you have cython installed, numpy (a numerical package for python), python (well... did I say this code was in python ?) with a version > 2.6. If this step fails again, kindly ask your system admin, (s)he is there for this, after all. Note that the installation (last command) is not strictly speaking mandatory.
Remember that if you modify CLASS to implement some new physics, you will need to perform this part again for the new CLASS.
The release of the Planck data comes with a likelihood program, called Clik, that one can recover from the ESA website, along with the data. Download all tar.gz files, extract them to the place of your convenience.
The Planck Likelihood Code (plc) is based on a library called clik. It will be extracted, alongside several .clik folders that contain the likelihoods. The installation of the code is described in the archive, and it uses an auto installer device, called waf.
Warning
Note that you are strongly advised to configure clik with the Intel mkl library, and not with lapack. There is a massive gain in execution time: without it, the code is dominated by the execution of the low-l polarisation data from WMAP.
Go to your plc folder, and execute the following line, taking into account the mkl installation
$ ./waf configure --install_all_deps --mkl=...
In your |MP| configuration file, to use this code, you should add the following line
path['clik'] = 'path/to/your/plc/folder/'The four likelihoods defined in |MP| for Planck are Planck_highl, Planck_lowl, Planck_lensing, lowlike (the polarization data from WMAP). In each of the respective data files for these likelihood, please make sure that the line, for instance,
Planck_highl.path_clik = data.path['clik']+'../something.clik'points to the correct clik file. Do not forget to source your Planck likelihood every time you want to use it:
$ source Your/Plc/bin/clik_profile.sh
You can put this line in your .bashrc file, and you should put it in your scripts for cluster computing.
Now the code is installed. Go anywhere, and just call
$ python montepython/MontePython.py --help $ python montepython/MontePython.py run --help $ python montepython/MontePython.py info --help
To see a list of all commands. For the run subcommand, there are two essential ones, without which the program will not start. At minimum, you should precise an output folder (-o) and a parameter file (-p). An example of parameter file is found in the main directory of MontePython (test.param, for instance).
A typical call would then be:
$ python montepython/MontePython.py run -o test -p example.param
If non existent, the test/ folder will be created, and a run with the number of steps described in example.param will be started. To run a chain with more steps, one can type:
$ python montepython/MontePython.py run -o test -p example.param -N 100
If you want to analyse the run, then just type
$ python montepython/MontePython.py info test/
Note that you probably want more than a hundred points before analyzing a folder.
Please refer to the pdf or online documentation for further details. The wiki contains additional details on installation. The forum also contains a collection of already answered questions, and can be used to discuss the code.
When using Monte Python in a publication, please acknowledge the code by citing the following papers. If you used Class, Nested Sampling or Cosmo Hammer, you should also cite the original works. Please also cite the relevant papers for each likelihood used: as of v3.0 we have a list of references for all likelihoods in the first of the papers below. Otherwise, this information can often be found in the .data file of the likelihood folder.
@article{Brinckmann:2018cvx,
author = "Brinckmann, Thejs and Lesgourgues, Julien",
title = "{MontePython 3: boosted MCMC sampler and other features}",
year = "2018",
eprint = "1804.07261",
archivePrefix = "arXiv",
primaryClass = "astro-ph.CO",
SLACcitation = "%%CITATION = ARXIV:1804.07261;%%"
}
@article{Audren:2012wb,
author = "Audren, Benjamin and Lesgourgues, Julien and Benabed,
Karim and Prunet, Simon",
title = "{Conservative Constraints on Early Cosmology: an
illustration of the Monte Python cosmological parameter
inference code}",
journal = "JCAP",
volume = "1302",
pages = "001",
doi = "10.1088/1475-7516/2013/02/001",
year = "2013",
eprint = "1210.7183",
archivePrefix = "arXiv",
primaryClass = "astro-ph.CO",
reportNumber = "CERN-PH-TH-2012-290, LAPTH-048-12",
SLACcitation = "%%CITATION = ARXIV:1210.7183;%%",
}