Dynamical decoupling: Support pulling through the whole circuit with non-clifford moments #6718

babacry · 2024-09-05T19:02:01Z

Previously, we only insert dd for split Clifford pieces. While, some use cases require

insert into consecutive idle moments cross Clifford pieces.
pull through clifford op even the moment contains non-clifford ops.
etc

Example: schema=(X, Y, X, Y)

Input:
                                                              ┌────────────────────────┐                             ┌──┐
0: ─────────────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@─────PhXZ(a=0.2,x=0.2,z=0.1)───H───
                                                               │                                                      │
1: ─────────────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@─────PhXZ(a=0.2,x=0.2,z=0.1)───H───

2: ───PhXZ(a=0.2,x=0.2,z=0.1)───@───PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────────H───
                                │                              │                                                      │
3: ─────────────────────────────┼───PhXZ(a=0.2,x=0.2,z=0.1)────┼──────────────────────@───────────────────────────────┼@──────────────────────────────H───
                                │                              │                      │                               ││
4: ─────────────────────────────┼──────────────────────────────┼──────────────────────@───────────────────────────────┼┼──────────────────────────────H───
                                │                              │                                                      ││
5: ───PhXZ(a=0.2,x=0.2,z=0.1)───@───PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@┼────PhXZ(a=0.2,x=0.2,z=0.1)───H───
                                                                                                                       │
6: ────────────────────────────────────────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)─────────────────────────────────@────PhXZ(a=0.2,x=0.2,z=0.1)───H───
                                                              └────────────────────────┘                             └──┘
Output:
                                                              ┌────────────────────────┐                             ┌──┐
0: ─────────────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@─────PhXZ(a=0.2,x=0.2,z=0.1)─────H────────────────────────
                                                               │                                                      │
1: ─────────────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@─────PhXZ(a=0.2,x=0.2,z=0.1)─────H────────────────────────

2: ───PhXZ(a=0.2,x=0.2,z=0.1)───@───PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@─────X───────────────────────────PhXZ(a=-0.5,x=0.5,z=0)───
                                │                              │                                                      │
3: ─────────────────────────────┼───PhXZ(a=0.2,x=0.2,z=0.1)────┼──────────────────────@────X──────────────────────────┼@────Y───────────────────────────H────────────────────────
                                │                              │                      │                               ││
4: ─────────────────────────────┼──────────────────────────────┼──────────────────────@────X──────────────────────────┼┼────Y───────────────────────────PhXZ(a=0.5,x=0.5,z=0)────
                                │                              │                                                      ││
5: ───PhXZ(a=0.2,x=0.2,z=0.1)───@───PhXZ(a=0.2,x=0.2,z=0.1)────@───────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)────@┼────PhXZ(a=0.2,x=0.2,z=0.1)─────H────────────────────────
                                                                                                                       │
6: ────────────────────────────────────────────────────────────PhXZ(a=0.2,x=0.2,z=0.1)─────X───────────────────────────@────PhXZ(a=-0.2,x=0.8,z=-0.5)───H────────────────────────
                                                              └────────────────────────┘                             └──┘
    """

codecov · 2024-09-05T19:15:45Z

Codecov Report

Attention: Patch coverage is 99.31034% with 1 line in your changes missing coverage. Please review.

Project coverage is 97.83%. Comparing base (bb24dae) to head (0b42839).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...irq-core/cirq/transformers/dynamical_decoupling.py	99.21%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6718      +/-   ##
==========================================
- Coverage   97.83%   97.83%   -0.01%     
==========================================
  Files        1077     1077              
  Lines       92537    92558      +21     
==========================================
+ Hits        90535    90555      +20     
- Misses       2002     2003       +1

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

eliottrosenberg

Looks good to me! Thank you so much, Renyi!

NoureldinYosri

my original concern still holds, the DD logic is getting more complex. can you simplify it?

cirq-core/cirq/transformers/dynamical_decoupling.py

NoureldinYosri · 2024-09-09T11:28:06Z

cirq-core/cirq/transformers/dynamical_decoupling.py

-            for qid, gate in enumerate(pulled_through_pauli_ops):
-                pulled_through *= gate.on(qubit_order[qid])
+        if _is_clifford_op(op_at_moment):
+            prev_circuit = cirq.Circuit(cirq.Moment(op_at_moment))


this is inefficient. looks like you only need the unitary which you can get using cirq.protocols.unitary_protocol.unitary(op)

Previously, we only insert dd for split Clifford pieces. While, one use case is to insert into consecutive idle moments cross Clifford pieces. We also support merging remaining Puali op of dd sequence into starting / ending single-qubit operations of each consecutive idle ops.

babacry · 2024-09-19T07:46:45Z

I have simplified the code with some abstractions.

Also added comments in a couple of places, especially pulling through mechanisms.

NoureldinYosri

a few nits, otherwise LGTM

NoureldinYosri · 2024-09-19T14:41:28Z

cirq-core/cirq/transformers/dynamical_decoupling.py

-            return False
-    return True
+def _is_clifford_op(op: 'cirq.Operation') -> bool:
+    if op.gate is not None and isinstance(op.gate, cirq.MeasurementGate):


Suggested change

if op.gate is not None and isinstance(op.gate, cirq.MeasurementGate):

if op.gate is not None or isinstance(op.gate, cirq.MeasurementGate):

cirq-core/cirq/transformers/dynamical_decoupling.py

NoureldinYosri

I just noticed you import cirq. this will cause circular imports. please follow the google style for imports

cirq-core/cirq/transformers/dynamical_decoupling.py

NoureldinYosri

thanks for the changes, great job

babacry · 2024-09-23T22:32:24Z

Thanks for the careful review Nour, will submit the code shortly.

eliottrosenberg · 2024-09-24T08:11:33Z

Thank you both!!

…non-clifford moments (quantumlib#6718) * Support a new feature of adding dd. Previously, we only insert dd for split Clifford pieces. While, one use case is to insert into consecutive idle moments cross Clifford pieces. We also support merging remaining Puali op of dd sequence into starting / ending single-qubit operations of each consecutive idle ops. * Pull through all the way. * Apply comments; simply the code a bit; Add comments. * apply reviews * import cirq -> import ops, ...

babacry requested review from vtomole and a team as code owners September 5, 2024 19:02

babacry requested a review from viathor September 5, 2024 19:02

babacry mentioned this pull request Sep 5, 2024

Implement dynamical decoupling transformer #6582

Open

babacry self-assigned this Sep 5, 2024

babacry requested review from eliottrosenberg and NoureldinYosri September 5, 2024 19:05

babacry force-pushed the dd branch 2 times, most recently from a1eb1ac to f571ccf Compare September 6, 2024 22:51

babacry changed the title ~~Support a new feature of adding dd.~~ Support a pulling through non-clifford moments in adding dd Sep 6, 2024

babacry changed the title ~~Support a pulling through non-clifford moments in adding dd~~ Dynamical decoupling: Support pulling through the whole circuit with non-clifford moments Sep 6, 2024

eliottrosenberg approved these changes Sep 7, 2024

View reviewed changes

NoureldinYosri requested changes Sep 9, 2024

View reviewed changes

CirqBot added the size: L 250< lines changed <1000 label Sep 10, 2024

babacry added 2 commits September 18, 2024 23:12

Pull through all the way.

a8be7c7

babacry force-pushed the dd branch 4 times, most recently from 08258a5 to 795bc43 Compare September 19, 2024 07:43

Apply comments; simply the code a bit; Add comments.

aa57a8b

babacry force-pushed the dd branch from 795bc43 to aa57a8b Compare September 19, 2024 07:53

babacry requested a review from NoureldinYosri September 19, 2024 08:00

NoureldinYosri reviewed Sep 19, 2024

View reviewed changes

apply reviews

e90191f

babacry force-pushed the dd branch from 0989397 to e90191f Compare September 20, 2024 21:44

babacry requested a review from NoureldinYosri September 20, 2024 21:51

NoureldinYosri requested changes Sep 21, 2024

View reviewed changes

cirq-core/cirq/transformers/dynamical_decoupling.py Outdated Show resolved Hide resolved

cirq-core/cirq/transformers/dynamical_decoupling.py Outdated Show resolved Hide resolved

cirq-core/cirq/transformers/dynamical_decoupling.py Show resolved Hide resolved

NoureldinYosri reviewed Sep 21, 2024

View reviewed changes

cirq-core/cirq/transformers/dynamical_decoupling.py Outdated Show resolved Hide resolved

NoureldinYosri reviewed Sep 21, 2024

View reviewed changes

cirq-core/cirq/transformers/dynamical_decoupling.py Outdated Show resolved Hide resolved

NoureldinYosri approved these changes Sep 23, 2024

View reviewed changes

import cirq -> import ops, ...

c6a2d3e

babacry force-pushed the dd branch from d256138 to c6a2d3e Compare September 23, 2024 21:56

babacry enabled auto-merge (squash) September 23, 2024 22:52

Merge branch 'main' into dd

0b42839

babacry merged commit 756abe3 into quantumlib:main Sep 24, 2024
35 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Dynamical decoupling: Support pulling through the whole circuit with non-clifford moments #6718

Dynamical decoupling: Support pulling through the whole circuit with non-clifford moments #6718

babacry commented Sep 5, 2024 •

edited

Loading

codecov bot commented Sep 5, 2024 •

edited

Loading

eliottrosenberg left a comment

NoureldinYosri left a comment

NoureldinYosri Sep 9, 2024

babacry commented Sep 19, 2024

NoureldinYosri left a comment

NoureldinYosri Sep 19, 2024

NoureldinYosri left a comment

NoureldinYosri left a comment

babacry commented Sep 23, 2024

eliottrosenberg commented Sep 24, 2024

	if op.gate is not None and isinstance(op.gate, cirq.MeasurementGate):
	if op.gate is not None or isinstance(op.gate, cirq.MeasurementGate):

Dynamical decoupling: Support pulling through the whole circuit with non-clifford moments #6718

Dynamical decoupling: Support pulling through the whole circuit with non-clifford moments #6718

Conversation

babacry commented Sep 5, 2024 • edited Loading

codecov bot commented Sep 5, 2024 • edited Loading

Codecov Report

eliottrosenberg left a comment

Choose a reason for hiding this comment

NoureldinYosri left a comment

Choose a reason for hiding this comment

NoureldinYosri Sep 9, 2024

Choose a reason for hiding this comment

babacry commented Sep 19, 2024

NoureldinYosri left a comment

Choose a reason for hiding this comment

NoureldinYosri Sep 19, 2024

Choose a reason for hiding this comment

NoureldinYosri left a comment

Choose a reason for hiding this comment

NoureldinYosri left a comment

Choose a reason for hiding this comment

babacry commented Sep 23, 2024

eliottrosenberg commented Sep 24, 2024

babacry commented Sep 5, 2024 •

edited

Loading

codecov bot commented Sep 5, 2024 •

edited

Loading