A general, minimal Python framework for building hybrid asynchronous decomposition samplers for quadratic unconstrained binary optimization (QUBO) problems.
dwave-hybrid facilitates three aspects of solution development:
- Hybrid approaches to combining quantum and classical compute resources
- Evaluating a portfolio of algorithmic components and problem-decomposition strategies
- Experimenting with workflow structures and parameters to obtain the best application results
The framework enables rapid development and insight into expected performance of productized versions of its experimental prototypes.
Your optimized algorithmic components and other contributions to this project are welcome!
Install from a package on PyPI:
pip install dwave-hybrid
or from source in development mode:
git clone https://github.com/dwavesystems/dwave-hybrid.git cd dwave-hybrid pip install -e .
Install test requirements and run unittest:
pip install -r tests/requirements.txt python -m unittest
import dimod
import hybrid
# Construct a problem
bqm = dimod.BinaryQuadraticModel({}, {'ab': 1, 'bc': -1, 'ca': 1}, 0, dimod.SPIN)
# Define the workflow
iteration = hybrid.RacingBranches(
hybrid.InterruptableTabuSampler(),
hybrid.EnergyImpactDecomposer(size=2)
| hybrid.QPUSubproblemAutoEmbeddingSampler()
| hybrid.SplatComposer()
) | hybrid.ArgMin()
workflow = hybrid.LoopUntilNoImprovement(iteration, convergence=3)
# Solve the problem
init_state = hybrid.State.from_problem(bqm)
final_state = workflow.run(init_state).result()
# Print results
print("Solution: sample={.samples.first}".format(final_state))Documentation for latest stable release included in Ocean is available as part of Ocean docs.
Released under the Apache License 2.0. See LICENSE file.
Ocean's contributing guide has guidelines for contributing to Ocean packages.