Mq Qaia Solver

What it does

When your agent uses it

• The user mentions QAIA
• Simulated bifurcation
• A combinatorial problem already encoded as Ising/QUBO

Example prompts

• “/mq-qaia-solver”

Requirements

• Python 3

Workflow steps

4 steps, taken from the first numbered list in SKILL.md.

1. 1In-place modification: solver.x is modified during update(). Pass x.copy() if you need the original.
2. 2Sparse matrices: Use scipy.sparse for large graphs — solvers accept both dense and sparse formats.
3. 3Symmetry: $J$ must be symmetric. If your adjacency matrix is asymmetric, symmetrize: J = (J + J.T) / 2.
4. 4Sign convention: QAIA energy uses $H = -\frac{1}{2}\sum_{i,j} J_{ij}s_is_j - \sum_i h_i s_i$. For Max-Cut examples in this repository…

What it can do on your machine

Read from SKILL.md and the folder at commit 2a0ca08. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are python).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Mq Qaia Solver loads about 2.2k tokens when it runs. Until then it costs about 133 tokens; SKILL.md has 604 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~133

When it runs · the whole SKILL.md, loaded when a task matches

~2.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from mindspore-ai/mindquantum at commit 2a0ca08, republished under its Apache-2.0 licence (© mindspore-ai). 604 words, ~2,231 tokens.

name

mq-qaia-solver

description

Solve Ising/QUBO-style combinatorial optimization problems using MindQuantum's Quantum Annealing-Inspired Algorithms (QAIA). Covers SimCIM, ASB/BSB/DSB, TSB/USB/LSB, LQA, CFC, CAC, SFC, and NMFA, with solver-specific CPU/GPU/NPU backend support. This is not circuit-based QAOA: QAIA solvers operate directly on an Ising coupling matrix J and optional field h. Use when the user mentions QAIA, SimCIM, simulated bifurcation, Ising solver, Max-Cut solver, QUBO, or a combinatorial problem already encoded as Ising/QUBO.

import numpy as np
from scipy.sparse import coo_matrix
from mindquantum.algorithm.qaia import BSB

# Define coupling matrix J (symmetric, from graph edges)
edges = [(0, 1), (1, 2), (2, 3), (3, 0), (0, 2)]
n_nodes = 4
row = [e[0] for e in edges] + [e[1] for e in edges]
col = [e[1] for e in edges] + [e[0] for e in edges]
data = [-1] * len(row)
J = coo_matrix((data, (row, col)), shape=(n_nodes, n_nodes))

# Solve
solver = BSB(J, batch_size=100, n_iter=500)
solver.update()

# Results
cuts = solver.calc_cut()
print(f"Best cut value: {max(cuts)}")
spins = np.sign(solver.x)  # Final spin configuration

solver = SolverClass(
    J,  # Coupling matrix: numpy.array or scipy.sparse
    h=None,  # External field: numpy.array [N, 1] (optional)
    x=None,  # Initial spins: numpy.array [N, batch_size] (optional)
    n_iter=1000,  # Number of iterations
    batch_size=1,  # Parallel samples
    backend="cpu-float32",  # Compute backend; supported values are solver-specific
)

solver.update()  # Run the optimization dynamics
solver.calc_cut()  # Max-Cut value for each batch (returns array)
solver.calc_energy()  # Ising energy for each batch
spins = np.sign(solver.x)  # Extract discrete spin assignments

solver = BSB(J, batch_size=1000, n_iter=2000, backend="gpu-float32")
solver.update()

import numpy as np
from scipy.sparse import coo_matrix
from mindquantum.algorithm.qaia import BSB, DSB, SimCIM


# 1. Load graph (e.g., GSet format)
def load_gset(filepath):
    """Load GSet benchmark graph as sparse coupling matrix."""
    import pandas as pd

    data = pd.read_csv(filepath, sep=" ", header=0)
    n = int(data.columns[0])
    rows = np.concatenate([data.iloc[:, 0] - 1, data.iloc[:, 1] - 1])
    cols = np.concatenate([data.iloc[:, 1] - 1, data.iloc[:, 0] - 1])
    vals = np.concatenate([data.iloc[:, 2], data.iloc[:, 2]])
    return coo_matrix((-vals, (rows, cols)), shape=(n, n))


# 2. Try multiple solvers
J = load_gset("G22.txt")  # 2000-node graph

results = {}
for name, Solver in [("BSB", BSB), ("DSB", DSB), ("SimCIM", SimCIM)]:
    solver = Solver(J, batch_size=100, n_iter=1000)
    solver.update()
    best_cut = max(solver.calc_cut())
    results[name] = best_cut
    print(f"{name}: MaxCut = {best_cut}")

# 3. Top result in this run
top = max(results, key=results.get)
print(f"Top solver in this run: {top} with cut = {results[top]}")

# J[i,j] = -weight(i,j) for edges, 0 otherwise
# h = None (no external field)

def qubo_to_ising(Q):
    """Convert upper-triangular QUBO matrix Q to QAIA Ising (J, h, constant)."""
    n = Q.shape[0]
    J = np.zeros((n, n))
    h = np.zeros(n)
    constant = 0.0
    for i in range(n):
        h[i] -= Q[i, i] / 2
        constant += Q[i, i] / 2
        for j in range(i + 1, n):
            qij = Q[i, j]
            J[i, j] = -qij / 4
            J[j, i] = -qij / 4
            h[i] -= qij / 4
            h[j] -= qij / 4
            constant += qij / 4
    return J, h.reshape(-1, 1), constant

© mindspore-ai, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/mq-qaia-solver of mindspore-ai/mindquantum.

Open the folder on GitHubat commit 2a0ca08