Kaiwu SDK Enterprise

Kaiwu SDK Enterprise is a Python SDK for building and solving QUBO and Ising optimization problems with Kaiwu classical optimizers and CIM services.

This package is the enterprise distribution. A valid license is required before using licensed features.

Requirements

• Python >= 3.10
• Windows, Linux, or macOS

Installation

• Python 3.10 https://www.python.org/downloads/release/python-31011/ 。
• pip3 install -i https://mirrors.aliyun.com/pypi/simple -r requirements/devel.txt

pip install kaiwu

Install a specific version:

pip install kaiwu==1.4.3

Quick Start

Solve an Ising Matrix

import numpy as np
import kaiwu as kw

matrix = -np.array(
    [
        [0.0, 1.0, 0.0, 1.0, 1.0],
        [1.0, 0.0, 0.0, 1.0, 1.0],
        [0.0, 0.0, 0.0, 1.0, 1.0],
        [1.0, 1.0, 1.0, 0.0, 1.0],
        [1.0, 1.0, 1.0, 1.0, 0.0],
    ]
)

optimizer = kw.classical.SimulatedAnnealingOptimizer(
    initial_temperature=100,
    alpha=0.99,
    cutoff_temperature=0.001,
    iterations_per_t=10,
    size_limit=10,
)

solutions = optimizer.solve(matrix)
print(solutions)

Solve a QUBO Matrix

solve_qubo() accepts a QuboModel. If you already have a QUBO matrix, convert it first:

import numpy as np
import kaiwu as kw

qubo_matrix = -np.array(
    [
        [0.0, -1.0, 0.0, 1.0, 1.0],
        [-1.0, 0.0, 0.0, 1.0, 1.0],
        [0.0, 0.0, 0.0, 1.0, 1.0],
        [1.0, 1.0, 1.0, 0.0, 1.0],
        [1.0, 1.0, 1.0, 1.0, 0.0],
    ]
)

# QUBO matrix conversion uses the upper-triangular convention.
qubo_model = kw.core.qubo_matrix_to_qubo_model(np.triu(qubo_matrix))

optimizer = kw.classical.SimulatedAnnealingOptimizer(
    initial_temperature=100,
    alpha=0.99,
    cutoff_temperature=0.001,
    iterations_per_t=10,
    size_limit=10,
)

solution, objective = optimizer.solve_qubo(qubo_model)
print(solution)
print(objective)

Use CIMOptimizer

CIMOptimizer submits or queries CIM tasks. Set a checkpoint directory before creating the optimizer.

import numpy as np
import kaiwu as kw

kw.common.CheckpointManager.save_dir = "tmp"

qubo_matrix = np.array(
    [
        [0.0, 1.0, 0.0],
        [0.0, 0.0, -2.0],
        [0.0, 0.0, 0.0],
    ]
)
qubo_model = kw.core.qubo_matrix_to_qubo_model(np.triu(qubo_matrix))

optimizer = kw.cim.CIMOptimizer(
    task_name="qubo_cim_example",
    wait=False,
    task_mode=kw.cim.TaskMode.OPTIMIZATION,
)

solution, objective = optimizer.solve_qubo(qubo_model)
print(solution, objective)

When wait=False, unfinished tasks may return (None, None). Use wait=True if the program should wait for the task result.

Main Modules

• kaiwu.classical: classical optimizers, including simulated annealing, tabu search, and brute force search
• kaiwu.cim: CIM task optimizer and task modes
• kaiwu.preprocess: matrix precision analysis and precision reduction tools
• kaiwu.hybrid: hybrid optimization tools
• kaiwu.sampler: sampler implementations
• kaiwu.hobo: higher-order binary optimization tools

Logging

Enable debug logs when needed:

import kaiwu as kw

kw.common.set_log_level("DEBUG")