cuQAOA is a CUDA-based QAOA Simulator, which implements state vector simulation of QAOA using Fast Walsh–Hadamard Transform.

In this project, we implement CUDA programs with the help of Numba library. So in order to use this project, you need to have numba installed following Numba offical documentation, and make sure CUDA GPU support for Numba is enabled.

numpy
scipy
numba
networkx

A class that is initialized with an array-like problem Hamiltonian $H_C$, and cuQAOASim will try to find the highest energy of $H_C$ (or equivalently, the ground energy of $-H_C$).

After initialized, you can invoke __call__ method with two arrays $\beta,\gamma$ that computes the expectation of $-H_C$ when QAOA arguments are configured $\beta,\gamma$.

from cuQAOA import cuQAOASim

from networkx import nx

N = 12
G = nx.erdos_renyi_graph(n=N, p=0.8)

cq = cuQAOASim([sum(1 if s >> x & 1 != s >> y & 1 else -1 for x, y in G.edges) for s in range(1 << N)])

print('random argument qaoa result =', (len(G.edges) - cq([1, 2, 3, 4], [5, 6, 7, 8])) / 2)

A wrapper for solving maxcut with QAOA.

from cuQAOA.maxcut import solve_maxcut, GraphType

from networkx import nx

N = 12
G = nx.erdos_renyi_graph(n=N, p=0.8)

print('optimized argument qaoa result =', solve_maxcut(G, 4, gtype=GraphType.ERDOS_RENYI))