Speed up some over parameterized tests #5606
Conversation
dabacon
requested review from
wcourtney,
a team,
vtomole,
cduck and
verult
as code owners
| @@ -58,9 +58,12 @@ def _all_rotation_pairs(): | |||
| yield (px, flip_x), (pz, flip_z) | |||
|
|
|||
|
|
|||
| @functools.cache | |||
| def _all_clifford_gates(): | |||
There was a problem hiding this comment.
Line not touched but might as well add a return type annotation while we're at it.
There was a problem hiding this comment.
and it made me realize this likely doesn't work since List is not hashable....so changed to tuple!
|
Looks like |
|
Naive python question: Why do these functions have different return types? from typing import Generator, Tuple
def tuple_of_ints() -> Tuple[int, ...]:
return tuple(
i for i in range(2)
)
def generator_of_tuple_of_ints() -> Generator[int, None, None]:
return (
i for i in range(2)
) |
|
You might think |
| @@ -739,7 +740,7 @@ def test_multi_clifford_decompose_by_unitary(): | |||
| # Construct a random clifford gate: | |||
| n, num_ops = 5, 20 # because we relied on unitary cannot test large-scale qubits | |||
| gate_candidate = [cirq.X, cirq.Y, cirq.Z, cirq.H, cirq.S, cirq.CNOT, cirq.CZ] | |||
| for seed in range(100): | |||
| for seed in range(10): | |||
There was a problem hiding this comment.
I'd suggest to drop seeding here and replace it with np.random calls and a simple repetition.
The default numpy random generator is getting reproducibly, non-constantly seeded
in check/pytest. Over time there should be more randomness in its generated choices than with the 10 seeds here.
|
LGTM, but I have noticed some strange random seeding in the test with seeds from Perhaps a matter for separate PR. |
pavoljuhas
added
the
automerge
CirqBot
added
the
front_of_queue_automerge
No description provided.