Generalized code base for modeling problems.
The diffpy.srfit package provides the framework for building a global optimizer on the fly from components such as function calculators (that calculate different data spectra), regression algorithms and structure models. The software is capable of co-refinement using multiple information sources or models. It provides a uniform interface for various regression algorithms. The target function being optimized can be specified by the user according to the data available.
Within the diffpy.srfit framework, any parameter used in describing the structure of a material can be passed as a refinable variable to the global optimizer. Once parameters are declared as variables they can easily be turned "on" or "off", i.e. fixed or allowed to vary. Additionally, variables may be constrained to obey mathematical relationships with other parameters or variables used in the structural model. Restraints can be applied to variables, which adds a penalty to the refinement process commensurate with the deviation from the known value or range. The cost function can also be customized by the user. If the refinement contains multiple models, each model can have its own cost function which will be properly weighted and combined to obtain the total cost function. Additionally, diffpy.srfit is designed to be extensible, allowing the user to integrate external calculators to perform co-refinements with other techniques.
SrFit has tools for coherently combining known information about a material to derive other properties, in particular material structure. SrFit allows the customization and creation of structure representations, profile calculators, constraints, restraints and file input parsers. The customized pieces can be glued together within SrFit to optimize a structure, or other physically relevant information from one or more experimental profiles. Other known information about the system of interest can be included with arbitrarily complex constraints and restraints. In this way, the end user creates a customized fitting application that suits the problem to the available information.
The subpackages herein define various pieces of the SrFit framework. Developers are encouraged to work through the examples described in the documentation to learn how to use and customize the various parts of SrFit.
For more information about the diffpy.srfit library, please consult our online documentation.
If you use diffpy.srfit in a scientific publication, we would like you to cite this package as
P. Juhás, C. L. Farrow, X. Yang, K. R. Knox and S. J. L. Billinge, Complex modeling: a strategy and software program for combining multiple information sources to solve ill posed structure and nanostructure inverse problems, Acta Crystallogr. A 71, 562-568 (2015).
The preferred method is to use Miniconda Python and install from the "conda-forge" channel of Conda packages.
To add "conda-forge" to the conda channels, run the following in a terminal.
conda config --add channels conda-forge
We want to install our packages in a suitable conda environment.
The following creates and activates a new environment named diffpy.srfit_env
conda create -n diffpy.srfit_env diffpy.srfit conda activate diffpy.srfit_env
The output should print the latest version displayed on the badges above.
This will install the minimal diffpy.srfit installation. It will often be used as along with other packages for manipulating and computing crystal structures and so on. We also therefore recommend installing the following:
diffpy.structure- crystal structure container and parsers, https://github.com/diffpy/diffpy.structurepyobjcryst- Crystal and Molecule storage, rigid units, bond length and bond angle restraints, https://github.com/diffpy/pyobjcryst
Optimizations involving pair distribution functions PDF or bond valence sums require
diffpy.srreal- python library for PDF calculation, https://github.com/diffpy/diffpy.srreal
Optimizations involving small angle scattering or shape characteristic functions from the diffpy.srfit.sas module require
sas- module for calculation of P(R) in small-angle scattering from the SasView project, http://www.sasview.org
If the above does not work, you can use pip to download and install the latest release from
Python Package Index.
To install using pip into your diffpy.srfit_env environment, type
pip install diffpy.srfit
If you prefer to install from sources, after installing the dependencies, obtain the source archive from
GitHub. Once installed, cd into your diffpy.srfit directory
and run the following
pip install .
This package also provides command-line utilities. To check the software has been installed correctly, type
diffpy.srfit --version
You can also type the following command to verify the installation.
python -c "import diffpy.srfit; print(diffpy.srfit.__version__)"
To view the basic usage and available commands, type
diffpy.srfit -h
You may consult our online documentation for tutorials and API references.
If you see a bug or want to request a feature, please report it as an issue and/or submit a fix as a PR.
Feel free to fork the project and contribute. To install diffpy.srfit in a development mode, with its sources being directly used by Python rather than copied to a package directory, use the following in the root directory
pip install -e .
To ensure code quality and to prevent accidental commits into the default branch, please set up the use of our pre-commit hooks.
- Install pre-commit in your working environment by running
conda install pre-commit. - Initialize pre-commit (one time only)
pre-commit install.
Thereafter your code will be linted by black and isort and checked against flake8 before you can commit. If it fails by black or isort, just rerun and it should pass (black and isort will modify the files so should pass after they are modified). If the flake8 test fails please see the error messages and fix them manually before trying to commit again.
Improvements and fixes are always appreciated.
Before contributing, please read our Code of Conduct.
For more information on diffpy.srfit please visit the project web-page or email the maintainers Caden Myers(cjm2304@columbia.edu) and Simon J. L. Billinge(sbillinge@ucsb.edu).
diffpy.srfit is built and maintained with scikit-package.
The source code in observable.py was derived from the 1.0 version of the Caltech "Pyre" project.