Skip to content

Commit

Permalink
Merge pull request #1171 from qiboteam/refactor_seed
Browse files Browse the repository at this point in the history 
Refactor seed generation in `quantum_info.random_ensembles`
  • Loading branch information
@renatomello
renatomello authored Feb 3, 2024
2 parents 4f1268c + 4e8b201 commit 6c5223d
Show file tree
Hide file tree
Showing 2 changed files with 51 additions and 103 deletions.
145 changes: 49 additions & 96 deletions src/qibo/quantum_info/random_ensembles.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@
)


class _probability_distribution_sin(rv_continuous):
class _probability_distribution_sin(rv_continuous): # pragma: no cover
def _pdf(self, theta: float):
return 0.5 * np.sin(theta)

Expand Down Expand Up @@ -53,21 +53,7 @@ def uniform_sampling_U3(ngates: int, seed=None, backend=None):
elif ngates <= 0:
raise_error(ValueError, f"ngates must be non-negative, but it is {ngates}.")

if (
seed is not None
and not isinstance(seed, int)
and not isinstance(seed, np.random.Generator)
):
raise_error(
TypeError, "seed must be either type int or numpy.random.Generator."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()

local_state = (
np.random.default_rng(seed) if seed is None or isinstance(seed, int) else seed
)
backend, local_state = _set_backend_and_local_state(seed, backend)

sampler = _probability_distribution_sin(a=0, b=np.pi, seed=local_state)
phases = local_state.random((ngates, 3))
Expand Down Expand Up @@ -133,21 +119,7 @@ def random_gaussian_matrix(
if stddev is not None and stddev <= 0.0:
raise_error(ValueError, "stddev must be a positive float.")

if (
seed is not None
and not isinstance(seed, int)
and not isinstance(seed, np.random.Generator)
):
raise_error(
TypeError, "seed must be either type int or numpy.random.Generator."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()

local_state = (
np.random.default_rng(seed) if seed is None or isinstance(seed, int) else seed
)
backend, local_state = _set_backend_and_local_state(seed, backend)

dims = (dims, rank)

Expand Down Expand Up @@ -193,10 +165,9 @@ def random_hermitian(
if not isinstance(semidefinite, bool) or not isinstance(normalize, bool):
raise_error(TypeError, "semidefinite and normalize must be type bool.")

if backend is None: # pragma: no cover
backend = GlobalBackend()
backend, local_state = _set_backend_and_local_state(seed, backend)

matrix = random_gaussian_matrix(dims, dims, seed=seed, backend=backend)
matrix = random_gaussian_matrix(dims, dims, seed=local_state, backend=backend)

if semidefinite:
matrix = np.dot(np.transpose(np.conj(matrix)), matrix)
Expand Down Expand Up @@ -241,11 +212,10 @@ def random_unitary(dims: int, measure: Optional[str] = None, seed=None, backend=
if measure != "haar":
raise_error(ValueError, f"measure {measure} not implemented.")

if backend is None: # pragma: no cover
backend = GlobalBackend()
backend, local_state = _set_backend_and_local_state(seed, backend)

if measure == "haar":
unitary = random_gaussian_matrix(dims, dims, seed=seed, backend=backend)
unitary = random_gaussian_matrix(dims, dims, seed=local_state, backend=backend)
Q, R = np.linalg.qr(unitary)
D = np.diag(R)
D = D / np.abs(D)
Expand Down Expand Up @@ -357,15 +327,14 @@ def random_quantum_channel(
NotImplementedError, f"order {order} not implemented for measure {measure}."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()
backend, local_state = _set_backend_and_local_state(seed, backend)

if measure == "bcsz":
super_op = _super_op_from_bcsz_measure(
dims=dims, rank=rank, order=order, seed=seed, backend=backend
dims=dims, rank=rank, order=order, seed=local_state, backend=backend
)
else:
super_op = random_unitary(dims, measure, seed, backend)
super_op = random_unitary(dims, measure, local_state, backend)
super_op = vectorization(super_op, order=order, backend=backend)
super_op = np.outer(super_op, np.conj(super_op))

Expand Down Expand Up @@ -573,28 +542,28 @@ def random_density_matrix(
elif normalize is True and basis is None:
raise_error(ValueError, "normalize cannot be True when basis=None.")

if backend is None: # pragma: no cover
backend = GlobalBackend()
backend, local_state = _set_backend_and_local_state(seed, backend)

if metric == "hilbert-schmidt":
rank = None

if pure:
state = random_statevector(dims, seed=seed, backend=backend)
state = random_statevector(dims, seed=local_state, backend=backend)
state = np.outer(state, np.transpose(np.conj(state)))
else:
if metric in ["hilbert-schmidt", "ginibre"]:
state = random_gaussian_matrix(
dims, rank, mean=0, stddev=1, seed=seed, backend=backend
dims, rank, mean=0, stddev=1, seed=local_state, backend=backend
)
state = np.dot(state, np.transpose(np.conj(state)))
state = state / np.trace(state)
else:
nqubits = int(np.log2(dims))
state = backend.identity_density_matrix(nqubits, normalize=False)
state += random_unitary(dims, seed=seed, backend=backend)
state += random_unitary(dims, seed=local_state, backend=backend)
state = np.dot(
state, random_gaussian_matrix(dims, rank, seed=seed, backend=backend)
state,
random_gaussian_matrix(dims, rank, seed=local_state, backend=backend),
)
state = np.dot(state, np.transpose(np.conj(state)))
state = state / np.trace(state)
Expand Down Expand Up @@ -662,21 +631,7 @@ def random_clifford(
f"return_circuit must be type bool, but it is type {type(return_circuit)}.",
)

if (
seed is not None
and not isinstance(seed, int)
and not isinstance(seed, np.random.Generator)
):
raise_error(
TypeError, "seed must be either type int or numpy.random.Generator."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()

local_state = (
np.random.default_rng(seed) if seed is None or isinstance(seed, int) else seed
)
backend, local_state = _set_backend_and_local_state(seed, backend)

hadamards, permutations = _sample_from_quantum_mallows_distribution(
nqubits, local_state=local_state
Expand Down Expand Up @@ -755,7 +710,7 @@ def random_clifford(
depth=1,
return_circuit=True,
density_matrix=density_matrix,
seed=seed,
seed=local_state,
backend=backend,
),
)
Expand Down Expand Up @@ -857,21 +812,7 @@ def random_pauli(
"subset argument must be a subset of strings in the set ['I', 'X', 'Y', 'Z'].",
)

if (
seed is not None
and not isinstance(seed, int)
and not isinstance(seed, np.random.Generator)
):
raise_error(
TypeError, "seed must be either type int or numpy.random.Generator."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()

local_state = (
np.random.default_rng(seed) if seed is None or isinstance(seed, int) else seed
)
backend, local_state = _set_backend_and_local_state(seed, backend)

complete_set = {"I": gates.I, "X": gates.X, "Y": gates.Y, "Z": gates.Z}

Expand Down Expand Up @@ -974,12 +915,11 @@ def random_pauli_hamiltonian(
"when normalize=True, gap is = 1, thus max_eigenvalue must be > 1.",
)

if backend is None: # pragma: no cover
backend = GlobalBackend()
backend, local_state = _set_backend_and_local_state(seed, backend)

d = 2**nqubits

hamiltonian = random_hermitian(d, normalize=True, seed=seed, backend=backend)
hamiltonian = random_hermitian(d, normalize=True, seed=local_state, backend=backend)

eigenvalues, eigenvectors = np.linalg.eigh(hamiltonian)

Expand Down Expand Up @@ -1079,21 +1019,7 @@ def random_stochastic_matrix(
if max_iterations <= 0.0:
raise_error(ValueError, "max_iterations must be a positive int.")

if (
seed is not None
and not isinstance(seed, int)
and not isinstance(seed, np.random.Generator)
):
raise_error(
TypeError, "seed must be either type int or numpy.random.Generator."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()

local_state = (
np.random.default_rng(seed) if seed is None or isinstance(seed, int) else seed
)
backend, local_state = _set_backend_and_local_state(seed, backend)

if precision_tol is None:
precision_tol = PRECISION_TOL
Expand Down Expand Up @@ -1281,3 +1207,30 @@ def _super_op_from_bcsz_measure(dims: int, rank: int, order: str, seed, backend)
super_op = operator @ super_op @ operator

return super_op


def _set_backend_and_local_state(seed, backend):
if (
seed is not None
and not isinstance(seed, int)
and not isinstance(seed, np.random.Generator)
):
raise_error(
TypeError, "seed must be either type int or numpy.random.Generator."
)

if backend is None: # pragma: no cover
backend = GlobalBackend()

if seed is None or isinstance(seed, int):
if backend.__class__.__name__ in [
"CupyBackend",
"CuQuantumBackend",
]: # pragma: no cover
local_state = backend.np.random.default_rng(seed)
else:
local_state = np.random.default_rng(seed)
else:
local_state = seed

return backend, local_state
9 changes: 2 additions & 7 deletions tests/test_quantum_info_random.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -61,11 +61,6 @@ def test_uniform_sampling_U3(backend, seed):
backend.assert_allclose(expectation_values[0], expectation_values[1], atol=1e-1)
backend.assert_allclose(expectation_values[0], expectation_values[2], atol=1e-1)

# execution for coverage
sampler = _probability_distribution_sin(a=0, b=np.pi, seed=seed)
sampler.pdf(1)
sampler.cdf(1)


@pytest.mark.parametrize("seed", [None, 10, np.random.default_rng(10)])
def test_random_gaussian_matrix(backend, seed):
Expand Down Expand Up @@ -341,8 +336,8 @@ def test_random_clifford(backend, nqubits, return_circuit, density_matrix, seed)

result_single = matrices.Z @ matrices.H

result_two = np.kron(matrices.H, matrices.S) @ np.kron(matrices.S, matrices.Z)
result_two = np.kron(matrices.Z @ matrices.S, matrices.I) @ result_two
result_two = np.kron(matrices.H, matrices.S) @ np.kron(matrices.S, matrices.Y)
result_two = np.kron(matrices.S @ matrices.X, matrices.I) @ result_two
result_two = matrices.CNOT @ matrices.CZ @ result_two

result = result_single if nqubits == 1 else result_two
Expand Down

0 comments on commit 6c5223d

Please sign in to comment.