formqubo

Hybrid quantum-classical optimization: convert Pyomo MIP/MIQP models to QUBO for D-Wave quantum annealing, with classical polishing via IPOPT/GLPK/CBC.

Setup with uv

1. Install uv (if not already installed)

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Create and activate a virtual environment

uv venv
source .venv/bin/activate   # macOS/Linux
# .venv\Scripts\activate    # Windows

3. Install core dependencies

uv pip install -e .

This installs pyomo and dimod (the required packages for simulation via simulated annealing).

4. Install D-Wave hardware support (optional)

To solve on real D-Wave hardware or use LeapHybridSampler, install the dwave extra:

uv pip install -e ".[dwave]"

You will also need a D-Wave Leap account and API token — set it via:

export DWAVE_API_TOKEN=your_token_here

5. Install a classical solver for the polishing stage

hybridsolve.py uses a classical solver (IPOPT, GLPK, or CBC) to polish continuous variables after the quantum stage. Install at least one:

# GLPK (open source, handles LP/MIP)
conda install -c conda-forge glpk
# or on macOS:
brew install glpk

# IPOPT (open source, handles nonlinear)
conda install -c conda-forge ipopt

# CBC (open source, handles MIP)
conda install -c conda-forge coin-or-cbc

Usage

import pyomo.environ as pyo
from scripts.pyomo_to_qubo import PyomoToQUBO
from scripts.hybridsolve import HybridSolver

# Build a Pyomo model
model = pyo.ConcreteModel()
# ... define variables, constraints, objective ...

# Convert to QUBO
converter = PyomoToQUBO(model)
qubo, offset = converter.to_qubo()

# Or solve end-to-end with the hybrid pipeline
solver = HybridSolver()
result = solver.solve(model)

Dependencies

Package	Role	Required
pyomo	Optimization modeling	Yes
dimod	QUBO/BQM representation + simulated annealing	Yes
dwave-ocean-sdk	D-Wave hardware samplers (LeapHybridSampler)	Optional
IPOPT / GLPK / CBC	Classical polishing solver	One required for HybridSolver