Skip to content

Latest commit

 

History

History
66 lines (50 loc) · 2.39 KB

README.md

File metadata and controls

66 lines (50 loc) · 2.39 KB
Raw

PyNumero

PyNumero: A high-level Python framework for rapid development of nonlinear optimization algorithms without large sacrifices on computational performance.

PyNumero dramatically reduces the time required to prototype new NLP algorithms and parallel decomposition while minimizing the performance penalty.

PyNumero libraries

PyNumero relies on C/C++ extensions for expensive computing operations.

If you installed Pyomo using conda (from conda-forge), then you can obtain precompiled versions of the redistributable interfaces (pynumero_ASL) using conda. Through Pyomo 5.6.9 these libraries are available by installing the pynumero_libraries package from conda-forge. Beginning in Pyomo 5.7, the redistributable pynumero libraries are included in the pyomo conda-forge package.

If you are not using conda or want to build the nonredistributable interfaces, you can build the extensions locally one of three ways:

  1. By running the build.py Python script in this directory. This script will automatically drive the cmake build harness to compile the libraries and install them into your local Pyomo configuration directory. Cmake options may be specified in the command. For example,

    python build.py -DBUILD_ASL=ON

If you have compiled Ipopt, and you would like to link against the libraries built with Ipopt, you can. For example,

python build.py -DBUILD_ASL=ON -DBUILD_MA27=ON -DIPOPT_DIR=<path_to_ipopt_build>/lib/

If you do so, you will likely need to update an environment variable for the path to shared libraries. For example, on Linux,

export LD_LIBRARY_PATH=<path_to_ipopt_build>/lib/
  1. By running pyomo build-extensions. This will build all registered Pyomo binary extensions, including PyNumero (using the build.py script from option 1).
  2. By manually running cmake to build the libraries. You will need to ensure that the libraries are then installed into a location that Pyomo (and PyNumero) can find them (e.g., in the Pyomo configuration directory, or in a common system location, or in a location included in the LD_LIBRARY_PATH environment variable).

Prerequisites

  1. pynumero_ASL:

    • cmake
    • a C/C++ compiler
    • ASL library and headers (optionally, the build harness can automatically check out and build AMPL/MP from GitHub to obtain this library)

  2. pynumero_MA27:

    • cmake
    • a C/C++ compiler
    • MA27 library