K. Jałowiecki, M. Rams and B. Gardas, "Brute-forcing spin-glass problems with CUDA", in preparation.

Documentation: https://ising.readthedocs.io/en/latest/

Ising is an open source package to solve small but otherwise abritrary spin-glass Ising models using exhaustive (brute force) search. It can serve as an excellent tool for benchmarking other solvers or generating low energy spectra (desirable e.g. for machine learning related tasks). The package is compatible with *NIX systems (and in principle should work on Windows too). Ising supports parallel computation via OpenMP or GPU, provided it has been build with CUDA support.

If you are running Linux and are interested in CPU-only implementation, you can install Ising from Python Package Index.

pip install ising

For other installation options, including building with CUDA support, please visit the official documentation.

The below example finds 4 lowest energy states of the Ising model defined by

H(s₀, s₁, s₂) =  − 2s₀s₁ + 3s₁s₂ + 2.5s₂s₃ − s₀

import ising

graph = {(0, 1): 2, (1, 2): -3, (2, 3): 2.5, (0, 0): 1}

result = ising.search(graph, num_states=4)
print(result.energies)

For advanced usage, including GPU support and tweaking execution parameters see documentation.