Set default unitary precision np.complex64 #5426
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dabacon
changed the title
dabacon
added
BREAKING CHANGE
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
95-martin-orion
approved these changes
Jun 14, 2022
Comment on lines
598
to
599
| kak decomposition is calculated from the `interaction`s. T | ||
| Thisd etermines how close $k_x$ must be to π/4 to guarantee |
There was a problem hiding this comment.
T Thisd etermines -> This determines
| @@ -83,7 +83,7 @@ def test_decompose_two_qubit_interaction_into_two_b_gates(obj: Any): | |||
| desired_unitary = obj if isinstance(obj, np.ndarray) else cirq.unitary(obj) | |||
| for operation in circuit.all_operations(): | |||
| assert len(operation.qubits) < 2 or operation.gate == _B | |||
| assert cirq.approx_eq(cirq.unitary(circuit), desired_unitary, atol=1e-6) | |||
| np.testing.assert_allclose(cirq.unitary(circuit), desired_unitary, atol=3e-5) | |||
There was a problem hiding this comment.
Should this also get the "lose a lot of precision" notice?
| @@ -384,8 +384,10 @@ def _create_corrected_circuit( | |||
| implements `target_unitary`. | |||
| """ | |||
| # Get the local equivalents | |||
| # We use higher precision than normal to avoid nummerical instabilities. | |||
There was a problem hiding this comment.
nummerical -> numerical
dabacon
requested review from
ColemanCollins,
nakardo and
gmauricio
as code owners
CirqBot
pushed a commit
that referenced
this pull request
Jun 29, 2022
…final_state_vector (#5636) Rolls back change in #5426 that set the default precision on `Circuit#unitary` to `np.complex128`. Also sets the default on `Circuit#final_state_vector` to `np.complex128`. There was a deprecation warning that this was going to be downgraded to `np.complex64`,, but this will now remain `np.complex128`. This is breaking in that is someone explicitly passed `dtype=None` to this method it would now break. Rolls back #5426 #5616 #5537 #5535
rht
pushed a commit
to rht/Cirq
that referenced
this pull request
May 1, 2023
rht
pushed a commit
to rht/Cirq
that referenced
this pull request
May 1, 2023
…final_state_vector (quantumlib#5636) Rolls back change in quantumlib#5426 that set the default precision on `Circuit#unitary` to `np.complex128`. Also sets the default on `Circuit#final_state_vector` to `np.complex128`. There was a deprecation warning that this was going to be downgraded to `np.complex64`,, but this will now remain `np.complex128`. This is breaking in that is someone explicitly passed `dtype=None` to this method it would now break. Rolls back quantumlib#5426 quantumlib#5616 quantumlib#5537 quantumlib#5535
harry-phasecraft
pushed a commit
to PhaseCraft/Cirq
that referenced
this pull request
Oct 31, 2024
harry-phasecraft
pushed a commit
to PhaseCraft/Cirq
that referenced
this pull request
Oct 31, 2024
…final_state_vector (quantumlib#5636) Rolls back change in quantumlib#5426 that set the default precision on `Circuit#unitary` to `np.complex128`. Also sets the default on `Circuit#final_state_vector` to `np.complex128`. There was a deprecation warning that this was going to be downgraded to `np.complex64`,, but this will now remain `np.complex128`. This is breaking in that is someone explicitly passed `dtype=None` to this method it would now break. Rolls back quantumlib#5426 quantumlib#5616 quantumlib#5537 quantumlib#5535
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This changes the
unitarymethod onCircuitto return a numpy array that usesnp.complex64instead ofnp.complex128.This is a breaking change.
Some ways we could deprecate this, but they all have some serious issues
We could add a deprecation warning as was done with
final_state_vectorif no override is None. See Consistent final_state_vector #5267. However, unlike in that PR, if we do make this warn in tests, we will need to fix many 1000s of tests (which we will then want to revert), becausefinal_state_vectoris not widely used, butunitaryis all over our tests.We could just start warning about this in our release notes, but not change it until a set version update (say 0.16). We could submit this PR, or use it for a future PR to update the tests as necessary.
Some comments on this change:
Mostly the changes here are to update the precision in tests, since we only have a little over 7 digits of precision in the new switch most of the precision is set to 1e-6. However there are cases where we build larger circuits that have many rounding errors and so end up closer to 5e-6 or even on 1e-5.
This changes the default precision on kak_decompositions, which was a large source of many of the test failures.
I check this for flakes at 200 repetitions. In the process I found some interesting failures. These are currently here, but I moved them out to #5431 for separate review.
Fixes: #1030