def simulate_one_amp(G, gamma, beta):
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.ansatz_state()
print(composer.circuit)
sim = QtreeSimulator()
result = sim.simulate(composer.circuit)
print('QTensor 1 amp', result.data)
print('QTensor 1 prob', np.abs(result.data)**2)
def get_tw_costs(N):
G, gamma, beta = get_test_problem(N, p=3, d=3)
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.energy_expectation_lightcone(list(G.edges())[0])
tn = QtreeTensorNet.from_qtree_gates(composer.circuit)
opt = OrderingOptimizer()
peo, _ = opt.optimize(tn)
tw = opt.treewidth
mems, flops = tn.simulation_cost(peo)
print('Max memory=', max(mems), 'Total flops=', sum(flops),
'Treewidth=', tw)
return tw, max(mems), sum(flops)
def test_tamaki_trimming_opt():
G, gamma, beta = get_test_problem(34, p=3, d=3)
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.ansatz_state()
sim = FeynmanSimulator()
sim.optimizer = SlicesOptimizer
result = sim.simulate(composer.circuit, batch_vars=3, tw_bias=7)
print(result)
sim.optimizer = TamakiTrimSlicing
sim.opt_args = {'wait_time': 5}
result_tam = sim.simulate(composer.circuit, batch_vars=3, tw_bias=7)
print(result_tam)
assert np.allclose(result_tam, result)
def test_profiled(capsys):
G, gamma, beta = get_test_problem()
composer = QtreeQAOAComposer(graph=G, gamma=[np.pi / 3], beta=[np.pi / 4])
composer.ansatz_state()
print(composer.circuit)
backend = PerfNumpyBackend()
sim = QtreeSimulator(bucket_backend=backend)
result = sim.simulate(composer.circuit)
print("Profile results")
print(backend.gen_report())
qtree_amp = result
assert qtree_amp
def test_parallel_batched():
G, gamma, beta = get_test_problem(14, 3, d=4)
batch_vars = 3
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.ansatz_state()
sim = QtreeSimulator()
amp = sim.simulate(composer.circuit)
amps = sim.simulate_batch(composer.circuit, batch_vars=2)
print('ordinary qtree amp', amp)
print('ordinary qtree 2 amps', amps)
assert abs(amp - amps[0]) < 1e-6
sim = FeynmanSimulator()
result = sim.simulate(composer.circuit, batch_vars=batch_vars, tw_bias=7)
print(result)
batch_amps = 2**batch_vars
assert len(result) == batch_amps
assert abs(amp - result[0]) < 1e-6
def test_qiskit_convert():
G, gamma, beta = get_test_problem()
qiskit_com = QiskitQAOAComposer(
graph=G, gamma=[np.pi/3], beta=[np.pi/4])
qiskit_com.ansatz_state()
# Convert Qiskit circuit to Qtree circuit
n, qc = from_qiskit_circuit(qiskit_com.circuit)
sim = QtreeSimulator()
all_gates = sum(qc, [])
# Simulate converted circuit
first_amp_from_qiskit = sim.simulate(all_gates)
com = QtreeQAOAComposer(
graph=G, gamma=[np.pi/3], beta=[np.pi/4])
com.ansatz_state()
# Simulate same circuit but created by Qtree composer
first_amp_orig = sim.simulate(com.circuit)
assert np.allclose(*[np.abs(x) for x in (first_amp_from_qiskit, first_amp_orig)])
assert np.allclose(first_amp_from_qiskit, first_amp_orig)
def test_orderings():
opt = OrderingOptimizer()
tam = TamakiOptimizer(wait_time=5)
seed = 43
np.random.seed(seed)
for n in range(14, 45, 2):
p = 3
G, gamma, beta = get_test_problem(n, p=p)
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.ansatz_state()
tn = QtreeTensorNet.from_qtree_gates(composer.circuit)
peo, tn = opt.optimize(tn)
treewidth = opt.treewidth
print_row(n, p, seed, 'greedy', treewidth)
peo, tn = tam.optimize(tn)
treewidth = tam.treewidth
print_row(n, p, seed, 'tamaki', treewidth)
def test_qtree_smoke():
G, gamma, beta = get_test_problem()
composer = QtreeQAOAComposer(
graph=G, gamma=[np.pi/3], beta=[np.pi/4])
composer.ansatz_state()
print(composer.circuit)
assert composer.circuit
assert composer.n_qubits == G.number_of_nodes()
composer = QtreeQAOAComposer(
graph=G, gamma=[np.pi/3], beta=[np.pi/4])
composer.energy_expectation_lightcone(list(G.edges())[0])
print(composer.circuit)
assert composer.circuit
assert composer.n_qubits == G.number_of_nodes()
def test_qtree():
G, gamma, beta = get_test_problem()
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.ansatz_state()
print(composer.circuit)
sim = QtreeSimulator()
result = sim.simulate(composer.circuit)
print(result)
qtree_amp = result
composer = CirqQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.ansatz_state()
print(composer.circuit)
sim = CirqSimulator()
result = sim.simulate(composer.circuit)
print(result)
final_cirq = result.final_state
assert final_cirq[0] - qtree_amp < 1e-5
assert result
from qtensor.optimisation.Optimizer import OrderingOptimizer from qtensor.optimisation.TensorNet import QtreeTensorNet from qtensor import QtreeQAOAComposer from qtensor import QAOAQtreeSimulator from qtensor import PerfNumpyBackend from qtensor import QAOA_energy import networkx as nx import numpy as np G = nx.random_regular_graph(4, 20) gamma, beta = [np.pi / 3], [np.pi / 2] composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta) composer.ansatz_state() backend = PerfNumpyBackend(print=True) sim = QAOAQtreeSimulator(composer, bucket_backend=backend) sim.simulate(composer.circuit) print(sim.bucket_backend) print(backend._profile_results) print(backend.gen_report())
def add_two_nodes_to_leafs(graph):
""" Works in-place """
leaves = [n for n in graph.nodes() if graph.degree(n) <= 1]
n = graph.number_of_nodes()
for leaf in leaves:
graph.add_edges_from([(leaf, n + 1), (leaf, n + 2)])
n += 2
graph = nx.Graph()
graph.add_edges_from([(0, 1)])
for i in range(n):
add_two_nodes_to_leafs(graph)
return graph
D = 6
G = bethe_lattice(D)
for p in range(1, D + 1):
print(f'{p=}, {G.number_of_nodes()=}')
gamma, beta = [0.1] * p, [0.2] * p
composer = QtreeQAOAComposer(graph=G, gamma=gamma, beta=beta)
composer.energy_expectation_lightcone((0, 1))
tn = QtreeTensorNet.from_qtree_gates(composer.circuit)
print(f'{tn.get_line_graph().number_of_nodes()=}')
opt = OrderingOptimizer()
peo, tn = opt.optimize(tn)
treewidth = opt.treewidth
print(f"{treewidth=}")