Add AnalogDetuneQubit cirq google version #7430

BichengYing · 2025-06-20T20:09:48Z

In order to keep using unit-style input, I have hacked the resolve_parameters function a little bit. Assume all inputs should either be tunit value or sympy.Symbol. (I.e. no sympy expression is allowed). I have to add this since we don't have SymbolFunc or ExprFunc in cirq world.

codecov · 2025-06-23T16:55:40Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.70%. Comparing base (c9839bf) to head (9f6c862).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #7430   +/-   ##
=======================================
  Coverage   98.69%   98.70%           
=======================================
  Files        1112     1114    +2     
  Lines       98214    98301   +87     
=======================================
+ Hits        96937    97024   +87     
  Misses       1277     1277

☔ View full report in Codecov by Sentry.
📢 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.

dstrain115 · 2025-06-23T21:20:08Z

cirq-google/cirq_google/json_test_data/AnalogDetuneQubit.json

+    "w": 5,
+    "target_freq": 8,
+    "prev_freq": null,
+    "neighbor_coupler_g_dict": null,


Should we provide values for these parameters in the json test?

We should but The expression must eval correctly when only cirq, pandas as pd, numpy as np and sympy have been imported. Maybe it is good to add tunits as import dependency in future PR? wdyt?

Oh, I see. This is fine for now then.

I think it might be worth adding tunits to the import list eventually in the import setup for json, but that can be done later.

yes, add a TODO in the json_serialization_test.py to mark this taks.

dstrain115 · 2025-06-23T21:21:05Z

cirq-google/cirq_google/ops/analog_detune_gates.py

+
+@cirq.value_equality(approximate=True)
+class AnalogDetuneQubit(cirq.ops.Gate):
+    """A step function that steup a qubit to the freq according to analog model.


nit: avoid abbreviations in the docstring if possible (freq => frequency)

dstrain115 · 2025-06-23T21:22:39Z

cirq-google/cirq_google/ops/analog_detune_gates.py

+        |          |---length --|
+        |
+        --------------------------(calculated from previous qubit freq using analog model)
+    Because the step hold at amp with inifinite length. You need prev_freq to compensate


This sentence needs to be adjusted "Because the step hold at amp with inifinite length.". I am not quite sure what you mean. Is it "Because the step is held at amp with infinite length, you need prev_freq ..."
(Also infinite is a typo)

Sorry for the confusing. Behind it is DetuneStepUp logic, that DAC amp is always on).

Update it into this:

Note that the step is held at a constant amplitude indefinitely. This gate is typically used by concatenating multiple instances. To ensure the curve is continuous and avoids sudden jumps, you need prev_freq to compensate for the previous Detune gate.

cirq-google/cirq_google/ops/analog_detune_gates.py

dstrain115 · 2025-06-23T21:25:38Z

cirq-google/cirq_google/ops/analog_detune_gates.py

+        # A shortcut for value resolution to avoid tu.unit compare with float issue.
+        def _direct_symbol_replacement(x, resolver: cirq.ParamResolver):
+            if isinstance(x, sympy.Symbol):
+                value = resolver.param_dict.get(x.name, "__NOT_FOUND__")


nit: Should "NOT_FOUND" be a constant?

I would still make this a string constant.

cirq-google/cirq_google/ops/analog_detune_gates.py

dstrain115 · 2025-06-23T21:28:45Z

cirq-google/cirq_google/ops/analog_detune_gates.py

+
+    def __init__(
+        self,
+        length: ValueOrSymbol,


Optional: I wonder if it will be easier to just make it length_ns and float rather than fight against cirq's typing. Sweeps etc are all defined as floats.

It is an important question. Yes, it should be easier in that way. The main reason I keep units is because of user's behavior. Trond: can you chime in on this? (Doug: can you help to add Trond as reviewer? I cannot do it)

Sweeps etc are all defined as floats: this is not true for example cirq.Points("a", [1,2,3]*tu.ns) is commonly used in pyle.

I think we managed to make it so that all the sweeps are with units, and it feels like this could make the user's life easier! :)

@trondandersen

I mean that the type is defined as floats in cirq. It works with things other than floats some of the time.

maybe I misunderstand the comment. Which line of code are you referring to? Here the type is ValueOrSymbol, which is tu.Value | Symbol and type checking is okay.

What I mean is this https://quantumai.google/reference/python/cirq/TParamVal
cirq itself defines a parameter value as a float, which is used in sweeps all over:
https://github.com/quantumlib/Cirq/blob/v1.5.0/cirq-core/cirq/study/sweeps.py#L41

Not much of the code actually depends that this is a float, but some of it (like gate implementations) does, so we are always going to be fighting against the typing system if we are trying to support sweeps with units.

I see. Yeah, i also feel there is always a fighting with value versus float convention. I think for this analog_experiment one, we will not have code about sweep to check-in.

In cirq-google sweep, we have expanded the proto definition to support tunits.

Cirq/cirq-google/cirq_google/api/v2/run_context.proto

Lines 226 to 250 in 7fe741b

// A list of explicit values.

message Points {

// The values.

repeated float points = 1;

tunits.Value unit = 2;

}

// A range of evenly-spaced values.

//

// Example: if the first_point is 1.0, the last_point is 2.0 ,

// and the num_points is 5, thi corresponds to the points

// 1.0, 1.25, 1.5, 1.75, 2.0

message Linspace {

// The start of the range.

float first_point = 1;

// The end of the range.

float last_point = 2;

// The number of points in the range (including first and last). Must be

// greater than zero. If it is 1, the first_point and last_point must be

// the same.

int64 num_points = 3;

tunits.Value unit = 4;

}

So running the experiment with tunit is okay and we don't need to check-in code to fight with type system. But happy to find a middle ground if we do encounter type checking issues.

dstrain115 · 2025-06-25T16:50:16Z

cirq-google/cirq_google/json_test_data/AnalogDetuneQubit.json

+    "w": 5,
+    "target_freq": 8,
+    "prev_freq": null,
+    "neighbor_coupler_g_dict": null,


I think it might be worth adding tunits to the import list eventually in the import setup for json, but that can be done later.

dstrain115 · 2025-06-25T16:50:51Z

cirq-google/cirq_google/ops/analog_detune_gates.py

+        # A shortcut for value resolution to avoid tu.unit compare with float issue.
+        def _direct_symbol_replacement(x, resolver: cirq.ParamResolver):
+            if isinstance(x, sympy.Symbol):
+                value = resolver.param_dict.get(x.name, "__NOT_FOUND__")


I would still make this a string constant.

In order to keep using unit-style input, I have hacked the _resolve_parameters_ function a little bit. Assume all inputs should either be tunit value or sympy.Symbol. (I.e. no sympy expression is allowed). I have to add this since we don't have SymbolFunc or ExprFunc in cirq world.

Add AnalogDetuneQubit cirq google version

55e7b2c

github-actions bot added the size: L 250< lines changed <1000 label Jun 20, 2025

BichengYing added 13 commits June 20, 2025 20:10

fix lint

5ef355c

fix lint

24ec27b

more fix

886e28a

type check

05abf57

Test coverage

fedea2f

fix json

8371b4b

fix

5406a87

repr

711e887

json

f84ec7a

remove comma

ba57ed4

json

bc8f5b2

coverage

a195673

coverage

98f487a

BichengYing marked this pull request as ready for review June 21, 2025 04:51

BichengYing requested review from wcourtney, vtomole, verult and a team as code owners June 21, 2025 04:51

BichengYing requested review from dstrain115 and eliottrosenberg June 21, 2025 04:53

BichengYing added 2 commits June 23, 2025 09:13

Merge branch 'main' into u/ybc/analog_detune

26b85b2

coverage

7668e7c

dstrain115 reviewed Jun 23, 2025

View reviewed changes

Update docstring

0975b7c

BichengYing requested a review from dstrain115 June 24, 2025 20:05

Merge branch 'main' into u/ybc/analog_detune

94db3e6

dstrain115 approved these changes Jun 25, 2025

View reviewed changes

BichengYing added 2 commits June 25, 2025 17:43

comment

e958382

Merge branch 'main' into u/ybc/analog_detune

9f6c862

BichengYing added this pull request to the merge queue Jun 25, 2025

Merged via the queue into quantumlib:main with commit 66b582f Jun 25, 2025
94 of 97 checks passed

BichengYing deleted the u/ybc/analog_detune branch June 25, 2025 20:28

This was referenced Aug 5, 2025

Add a new transformer that performs random pauli insertion #7558

Open

update the RB fitting method to compensate for the standard deviation induced from the random number generator #7562

Merged

	// A list of explicit values.
	message Points {
	// The values.
	repeated float points = 1;

	tunits.Value unit = 2;
	}

	// A range of evenly-spaced values.
	//
	// Example: if the first_point is 1.0, the last_point is 2.0 ,
	// and the num_points is 5, thi corresponds to the points
	// 1.0, 1.25, 1.5, 1.75, 2.0
	message Linspace {
	// The start of the range.
	float first_point = 1;
	// The end of the range.
	float last_point = 2;
	// The number of points in the range (including first and last). Must be
	// greater than zero. If it is 1, the first_point and last_point must be
	// the same.
	int64 num_points = 3;

	tunits.Value unit = 4;
	}

Add AnalogDetuneQubit cirq google version #7430

Add AnalogDetuneQubit cirq google version #7430

Uh oh!

Conversation

BichengYing commented Jun 20, 2025

Uh oh!

codecov bot commented Jun 23, 2025 • 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!

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!

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!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

BichengYing Jun 23, 2025 • 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!

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!

Uh oh!

Uh oh!

codecov bot commented Jun 23, 2025 •

edited

Loading

BichengYing Jun 23, 2025 •

edited

Loading