Migrate the remainder of cirq-core to np.random.Generators by arettig · Pull Request #8527 · quantumlib/Cirq

arettig · 2026-11-11T21:15:40Z

Part of #8374

This replaces the usage of parse_random_state with the new parse_random_generator in cirq-core. Several tests required specific seeds to pass and these were fixed in several ways:

In cases where the test simply asserted that a given seed gave an expected output, the test was changed to assert that the same seed always produces the same output by running multiple times. One assertion to an expected output was kept to cover portability.
In some cases, specific seeds were found that yielded random unitaries with a desired property for a test - e.g. were in a certain portion of the Weyl chamber. These tests were fixed by simply searching for new seeds. Ideally these tests should not rely on randomness for this, but that can be fixed in a separate PR.
In some tests, results were compared too strictly for the statistical uncertainty, which passed previously because the seed was fixed. In these cases, the number of samples was increased to tighten the uncertainty sufficiently to avoid flakiness.

Potentially breaking behavior:

Outputs for a fixed seed will not reproduce results obtained before this PR.

Replaces the use of RandomState in the random circuit generators with numpy Generators. Replaces cases where the tests depended on the random seeding/algorithm.

codecov · 2026-11-11T21:25:47Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.58%. Comparing base (8a8fc5c) to head (80a317a).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8527      +/-   ##
==========================================
- Coverage   99.58%   99.58%   -0.01%     
==========================================
  Files        1137     1137              
  Lines      105119   105133      +14     
==========================================
+ Hits       104687   104698      +11     
- Misses        432      435       +3

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

pavoljuhas

LGTM after addressing some small issues (see inline) and a few leftover references to np.random.RandomState in touched files.

Thank you for this conversion!

pavoljuhas · 2026-11-12T06:59:40Z

    qubits = cirq.LineQubit.range(num_qubits)
    circuit = cirq.Circuit()
-    random_state = cirq.value.parse_random_state(random_state)
+    random_state = cirq.value.parse_random_generator(random_state)


Nit - should we rename LHS to prng here?

pavoljuhas · 2026-11-12T07:10:48Z

@@ -33,7 +33,7 @@ def get_seeded_initial_mapping(graph_seed, init_seed) -> dict[cirq.Qid, cirq.Qid

 @pytest.mark.parametrize('seed', [random.randint(0, 2**32) for _ in range(10)])


Nit - can we replace the above with one static sample, i.e., a fixed list of 10 integers generated as above?

This would let us have constant parameterized test names (which include parameter value) with effectively the same test.

pavoljuhas · 2026-11-12T22:10:00Z



 def test_benchmark_2q_xeb_fidelities_vectorized() -> None:
    rs = np.random.RandomState(52)


I think we can replace these instances of RandomState - the patch below seems to work:

diff --git a/cirq-core/cirq/experiments/xeb_fitting_test.py b/cirq-core/cirq/experiments/xeb_fitting_test.py index bd1fb45c9..363e0681c 100644 --- a/cirq-core/cirq/experiments/xeb_fitting_test.py +++ b/cirq-core/cirq/experiments/xeb_fitting_test.py @@ -142,6 +142,8 @@ def test_benchmark_2q_xeb_fidelities_parallel() -> None: def test_benchmark_2q_xeb_fidelities_vectorized() -> None: - rs = np.random.RandomState(52) - mock_records = [{'pure_probs': rs.rand(4), 'sampled_probs': rs.rand(4)} for _ in range(100)] + prng = cirq.value.parse_random_generator(52) + mock_records = [ + {'pure_probs': prng.random(4), 'sampled_probs': prng.random(4)} for _ in range(100) + ] df = pd.DataFrame(mock_records) @@ -307,7 +309,7 @@ def test_parallel_full_workflow(request, use_pool) -> None: def test_fit_exponential_decays() -> None: - rs = np.random.RandomState(999) + prng = cirq.value.parse_random_generator(999) cycle_depths = np.arange(3, 100, 11) - fidelities = 0.95 * 0.98**cycle_depths + rs.normal(0, 0.2) + fidelities = 0.95 * 0.98**cycle_depths + prng.normal(0, 0.02) a, layer_fid, a_std, layer_fid_std = _fit_exponential_decay(cycle_depths, fidelities) np.testing.assert_allclose([a, layer_fid], [0.95, 0.98], atol=0.02) @@ -317,7 +319,7 @@ def test_fit_exponential_decays() -> None: def test_fit_exponential_decays_negative_fids() -> None: - rs = np.random.RandomState(999) + prng = cirq.value.parse_random_generator(999) cycle_depths = np.arange(3, 100, 11) - fidelities = 0.5 * 0.5**cycle_depths + rs.normal(0, 0.2) - 0.5 + fidelities = 0.5 * 0.5**cycle_depths + prng.normal(0, 0.2) - 0.5 assert np.sum(fidelities > 0) <= 1, 'they go negative' a, layer_fid, a_std, layer_fid_std = _fit_exponential_decay(cycle_depths, fidelities)

pavoljuhas · 2026-11-12T22:20:31Z

+        random_state: A seed (int) or `np.random.Generator` to use when
+            generating random values. If not set, a new unseeded generator is
+            used.


Nit on wording - feel free to adjust:

Suggested change

random_state: A seed (int) or `np.random.Generator` to use when

generating random values. If not set, a new unseeded generator is

used.

random_state: A seed (int) or `np.random.Generator` to use when

generating random values. Use a default unseeded generator

when unspecified.

Also please check and sync docstrings of other functions here that take random_state.

pavoljuhas · 2026-11-12T22:29:23Z

+        rng: Random number generator (or seed) to be used in sampling. Anything
+            accepted by `cirq.value.parse_random_generator`. Default is a new,
+            unseeded `np.random.Generator`.


Nit:

Suggested change

rng: Random number generator (or seed) to be used in sampling. Anything

accepted by `cirq.value.parse_random_generator`. Default is a new,

unseeded `np.random.Generator`.

rng: Random number generator (or seed) to be used in sampling. Anything

accepted by `cirq.value.parse_random_generator`. If unspecified,

use a default unseeded `np.random.Generator`.

pavoljuhas · 2026-11-12T22:32:33Z

    """
-    real_rng = random_state.parse_random_state(rng)
+    real_rng = random_state.parse_random_generator(rng)
+    size = tuple(shape)


Nit - size can be confusing as it evokes a scalar. Let us just coerce shape or, if type checker complains, use a new variable tshape.

Suggested change

size = tuple(shape)

shape = tuple(shape)

pavoljuhas · 2026-11-12T22:37:09Z

    tab = two_qubit_gate_product_tabulation(
        base_gate, 0.1, sample_scaling=2, random_state=np.random.RandomState(seed)
    )


Suggested change

tab = two_qubit_gate_product_tabulation(base_gate, 0.1, sample_scaling=2, random_state=seed)

)

pavoljuhas · 2026-11-12T22:42:30Z



 def _cz_with_adjacent_z_rotations(
    a: cirq.GridQubit, b: cirq.GridQubit, prng: np.random.RandomState


Suggested change

a: cirq.GridQubit, b: cirq.GridQubit, prng: np.random.Generator

pavoljuhas · 2026-11-12T22:47:09Z



 def random_locals(x: float, y: float, z: float, seed=None) -> np.ndarray:
    rng = np.random.RandomState(seed)


rng = cirq.value.parse_random_generator(seed)

pavoljuhas · 2026-11-12T22:49:54Z



 def test_numerical_compiler_json_rejects_live_rng() -> None:
    compiler = TwoQubitNumericalCompiler(base_gates=(_CZ,), random_state=np.random.RandomState(1))


Suggested change

compiler = TwoQubitNumericalCompiler(base_gates=(_CZ,), random_state=np.random.default_rng())

arettig added 3 commits November 11, 2026 17:57

Convert random circuits to np generators

f0a647f

Replaces the use of RandomState in the random circuit generators with numpy Generators. Replaces cases where the tests depended on the random seeding/algorithm.

Remove seed hardcoding from some tests

604d539

Migrate to generators in contrib

4555969

github-actions Bot added the size: L 250< lines changed <1000 label Nov 11, 2026

Clean up a test

80a317a

arettig marked this pull request as ready for review November 11, 2026 21:55

arettig requested a review from a team as a code owner November 11, 2026 21:55

arettig requested review from fdmalone and pavoljuhas November 11, 2026 21:55

pavoljuhas requested changes Nov 12, 2026

View reviewed changes

Navigation Menu

Repository navigation

Migrate the remainder of cirq-core to np.random.Generators - #8527

Migrate the remainder of cirq-core to np.random.Generators#8527
arettig wants to merge 4 commits into
quantumlib:mainfrom
arettig:rsos_generator_pr3

arettig commented Nov 11, 2026

Uh oh!

codecov Bot commented Nov 11, 2026 •

edited

Loading

Uh oh!

pavoljuhas left a comment

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

pavoljuhas Nov 12, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		@@ -33,7 +33,7 @@ def get_seeded_initial_mapping(graph_seed, init_seed) -> dict[cirq.Qid, cirq.Qid

		@pytest.mark.parametrize('seed', [random.randint(0, 2**32) for _ in range(10)])



		def test_benchmark_2q_xeb_fidelities_vectorized() -> None:
		rs = np.random.RandomState(52)


	tab = two_qubit_gate_product_tabulation(base_gate, 0.1, sample_scaling=2, random_state=seed)
	)



		def _cz_with_adjacent_z_rotations(
		a: cirq.GridQubit, b: cirq.GridQubit, prng: np.random.RandomState



		def random_locals(x: float, y: float, z: float, seed=None) -> np.ndarray:
		rng = np.random.RandomState(seed)



		def test_numerical_compiler_json_rejects_live_rng() -> None:
		compiler = TwoQubitNumericalCompiler(base_gates=(_CZ,), random_state=np.random.RandomState(1))

Conversation

arettig commented Nov 11, 2026

Uh oh!

codecov Bot commented Nov 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

pavoljuhas left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

codecov Bot commented Nov 11, 2026 •

edited

Loading