Implement dynamical decoupling. by babacry · Pull Request #6515 · quantumlib/Cirq

babacry · 2024-03-20T21:23:05Z

Add dynamical decoupling operations to circuits for idle moments.

Two methods are provided to apply dynamical decoupling.

Schema "XX_PAIR" and "YY_PAIR" are provided.
Users can provide their own base dynamical decoupling sequence via base_dd_sequence.

We repeat given dd sequence in continuous idle moments until the remaining idle moments isn't long enough for another repetition.

babacry · 2024-03-20T21:25:35Z

Post the draft for general suggestions.

And see how the CI goes compared with local CI.

eliottrosenberg

LGTM up to some nits and suggestions. Once this is merged, we can discuss some possible ways to augment it. Thanks!

eliottrosenberg · 2024-03-25T20:33:02Z

+) -> list['cirq.Gate']:
+    match schema:
+        case _DynamicalDecouplingSchema.XX_PAIR:
+            return _repeat_sequence([cirq.XPowGate(), cirq.XPowGate()], num_idle_moments)


I think you want cirq.X instead of cirq.XPowGate().

eliottrosenberg · 2024-03-25T20:33:19Z

+        case _DynamicalDecouplingSchema.XX_PAIR:
+            return _repeat_sequence([cirq.XPowGate(), cirq.XPowGate()], num_idle_moments)
+        case _DynamicalDecouplingSchema.YY_PAIR:
+            return _repeat_sequence([cirq.YPowGate(), cirq.YPowGate()], num_idle_moments)


Similarly, I think you want cirq.Y here.

eliottrosenberg · 2024-03-25T20:34:49Z

+        raise ValueError('Invalid dynamical decoupling sequence. Expect more than one gates.')
+    matrices = [cirq.unitary(gate) for gate in dd_sequence]
+    product = reduce(np.matmul, matrices)
+    if not np.array_equal(product, np.eye(2)):


You could use cirq.equal_up_to_global_phase() here (since we don't care if the sequence differs from the identity by a global phase).

eliottrosenberg · 2024-03-25T20:39:20Z

+class _DynamicalDecouplingSchema(enum.Enum):
+    """Supported schemes of dynamical decoupling."""
+
+    XX_PAIR = 'XX_PAIR'


We may also want one that is does X X**-1. X and X**-1 have the same unitary but are implemented differently on hardware. Similarly for Y.

snichet

A couple suggestions for optimal implementation on hardware.

snichet · 2024-03-25T22:57:36Z

+        dd_model=DynamicalDecouplingModel.from_schema("XX_PAIR"),
+    )
+
+    # Insert one XX_PAIR dynamical decoupling sequence in idle moments.


should be YY_Pair

snichet · 2024-03-25T23:02:55Z

+        return self.schema, self.base_dd_sequence
+
+
+def add_dynamical_decoupling(


Might be nice to have an option only to insert DD gates on moments that only have single qubits gates. On the hardware, the calibrations for the gates do not include calibrating single qubits and two qubits gates concurrently.

Good idea, tracking this here. Feel free to put more thoughts there, thanks!

snichet · 2024-03-25T23:22:55Z

+                num_idle_moments=moment_id - last_busy_moment_by_qubits[q] - 1
+            )
+            for idx, gate in enumerate(insert_gates):
+                insert_into.append((last_busy_moment_by_qubits[q] + idx + 1, gate.on(q)))


It seems like this is only inserting X gates on consecutive idle moments. I think the next step would be to divide the circuit into moments that are Clifford, then insert X gates on all idle single qubit moments (and perhaps also idle 2 qubit moments, but the caveat is my comment above). Then use the fact that Clifford circuits can be quickly and efficiently simulated to put the correct Clifford gate at the end of the group of moments, to inverse all the previous inserted gates. For CZ gates, this should just be a X, Y or Z gate, i believe.

Good idea, tracking this here. Feel free to put more thoughts there, thanks!

codecov · 2024-05-01T20:59:57Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.79%. Comparing base (bfba965) to head (b9c9557).
Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6515      +/-   ##
==========================================
- Coverage   97.79%   97.79%   -0.01%     
==========================================
  Files        1124     1126       +2     
  Lines       95705    95784      +79     
==========================================
+ Hits        93595    93670      +75     
- Misses       2110     2114       +4

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

NoureldinYosri · 2024-05-02T20:27:27Z

+        )
+
+
+@value.value_equality


it doesn't look like we mutate the properties of this class, consider making this a dataclass/attrs.frozen e.g.

@attrs.frozen class DynamicalDecouplingModel: schema: Optional[_DynamicalDecouplingSchema] = attrs.field(default=None, validator=...) base_dd_sequence: Optional[list['cirq.Gate']] = attrs.field(default=None, validator=...) ...etc

actually I don't think this class is needed.. consider removing it, this will simplify your design and make your life easier as you won't need to worry about json

Done.

Context: I was not so sure about whether to add a model for dynamical decoupling. I chose to add a model class as I expected the dynamical decoupling schema might get more and more complex in the future. While for now, we only need to handle the case of "Supported schemes" and "Customized sequence". A function is good enough, and we do not need to worry about json.

NoureldinYosri · 2024-05-02T20:36:07Z

+    circuit: 'cirq.AbstractCircuit',
+    *,
+    context: Optional['cirq.TransformerContext'] = None,
+    dd_model: DynamicalDecouplingModel = DynamicalDecouplingModel.from_schema("X_XINV"),


instead of dd_model try

spin_echo_sequence: Sequence[cirq.Gate] = (cirq.X, cirq.X),

Updated with schema.

My understanding is "dynamical decoupling sequence" is more general than "spin echo sequence", correct me if I was wrong @eliottrosenberg .

NoureldinYosri · 2024-05-02T20:42:32Z

+    dd_model: DynamicalDecouplingModel = DynamicalDecouplingModel.from_schema("X_XINV"),
+) -> 'cirq.Circuit':
+    """Add dynamical decoupling gate operations to a given circuit.
+


does this preserve the moment structure? consider adding this to the docstring.

It preserves the moment structure.

Updated docstring.

NoureldinYosri

just a little bit more

NoureldinYosri · 2024-05-09T20:56:22Z

+        except ValueError:
+            raise
+    else:
+        is_valid, error_message = _validate_dd_sequence(schema)


this is a golang style, prefer to just raise the error inside _validate_dd_sequence

NoureldinYosri · 2024-05-09T20:57:14Z

+    last_busy_moment_by_qubits: Dict['cirq.Qid', int] = {q: 0 for q in circuit.all_qubits()}
+    insert_into: list[Tuple[int, 'cirq.OP_TREE']] = []
+
+    if isinstance(schema, str):


consider grouping this logic into a function (e.g. _parse_dd_sequence(schema))

NoureldinYosri

@babacry congratulations on your first PR

babacry · 2024-05-10T00:00:31Z

Thanks for the careful review, everyone!

babacry requested review from a team, cduck and vtomole as code owners March 20, 2024 21:23

babacry requested a review from viathor March 20, 2024 21:23

babacry marked this pull request as draft March 20, 2024 21:23

NoureldinYosri self-assigned this Mar 20, 2024

NoureldinYosri self-requested a review March 20, 2024 21:26

babacry force-pushed the dd branch from eabbb63 to 74e1df0 Compare March 20, 2024 23:26

eliottrosenberg approved these changes Mar 25, 2024

View reviewed changes

snichet reviewed Mar 25, 2024

View reviewed changes

Impl dynamical decoupling transfomer.

6e65ad8

babacry force-pushed the dd branch from 56d9010 to 856583a Compare May 1, 2024 19:32

Address comments.

3a6a9b4

babacry force-pushed the dd branch from 856583a to 3a6a9b4 Compare May 1, 2024 19:34

Merge branch 'quantumlib:main' into dd

427227d

babacry mentioned this pull request May 1, 2024

Implement dynamical decoupling transformer #6582

Closed

babacry marked this pull request as ready for review May 1, 2024 20:30

NoureldinYosri requested changes May 2, 2024

View reviewed changes

CirqBot added the size: L 250< lines changed <1000 label May 5, 2024

babacry force-pushed the dd branch from cb5e344 to bd366af Compare May 8, 2024 03:10

Apply comments of removing dynamical decoupling model class.

3ae49b1

babacry force-pushed the dd branch from bd366af to 3ae49b1 Compare May 8, 2024 03:15

NoureldinYosri reviewed May 9, 2024

View reviewed changes

babacry and others added 3 commits May 9, 2024 14:35

Apply comments.

c2e8026

Merge branch 'main' into dd

e30787a

Merge branch 'main' into dd

b9c9557

NoureldinYosri approved these changes May 9, 2024

View reviewed changes

NoureldinYosri merged commit f79f902 into quantumlib:main May 9, 2024

jselig-rigetti pushed a commit to jselig-rigetti/Cirq that referenced this pull request May 28, 2024

Implement dynamical decoupling. (quantumlib#6515)

77e3a0f

harry-phasecraft pushed a commit to PhaseCraft/Cirq that referenced this pull request Oct 31, 2024

Implement dynamical decoupling. (quantumlib#6515)

977594d

BichengYing pushed a commit to BichengYing/Cirq that referenced this pull request Jun 20, 2025

Implement dynamical decoupling. (quantumlib#6515)

39cd221

		return self.schema, self.base_dd_sequence


		def add_dynamical_decoupling(

		)


		@value.value_equality

Conversation

babacry commented Mar 20, 2024

Uh oh!

babacry commented Mar 20, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

eliottrosenberg 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!

snichet 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!

codecov bot commented May 1, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

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!

babacry May 8, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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!

NoureldinYosri 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!

NoureldinYosri left a comment

Choose a reason for hiding this comment

Uh oh!

babacry commented May 10, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

babacry commented Mar 20, 2024 •

edited

Loading

codecov bot commented May 1, 2024 •

edited

Loading

babacry May 8, 2024 •

edited

Loading