RB utils class #86

gadial · 2021-06-07T06:23:07Z

Summary

Details and comments

nkanazawa1989 · 2021-06-11T03:52:56Z

qiskit_experiments/randomized_benchmarking/rb_utils.py

+from qiskit.providers.backend import Backend
+
+
+class RBUtils:


Why this need to be a class? Looks like this is collection of functions.

We might change it since it may be better to put those functions inside of the experiment/analysis classes. So far I'm trying to get the Ignis format to work before committing to more structural changes.

You mean adding them to RBAnalysis as methods? I think this means all RB subclasses inherit those functionality, even though they don't use, i.e. IRB doesn't need to calculate EPG because we can directly estimate it.

…g machine

ShellyGarion

Generally, the code LGTM! But it's difficult to verify that the calculations are correct.
I would suggest to add some examples of EPG and coherence limit calculations to the tutorial like it's done in:
https://qiskit.org/documentation/tutorials/noise/4_randomized_benchmarking.html

…g machine

nkanazawa1989

Hi Gadi, thanks for adding the utils. I'm happy to see we can reuse resources from Ignis. I added some suggestions for handling of options.

nkanazawa1989 · 2021-06-18T02:24:12Z

qiskit_experiments/base_experiment.py

        """
        self._analysis_options.update_options(**fields)
+
+    def _postprocess_transpiled_circuits(self, circuits: List[QuantumCircuit]):


Post processing is quite limited context and many experiments don't use though we are adding this to the base class. I think returning transpile method back (and calling this method from .run, instead of just calling terra transpiler) would help other experiments too, i.e. in QV experiment people may want override transpiler. @chriseclectic ?

My goal here is to minimize as much as possible taking over processes which the base class should handle. I don't need control over the transpilation stage itself; only a peek into its results (similar to the post processing in use in the analysis class). Of course, I'm happy to hear any suggestion you or @chriseclectic might have.

I think this sort of step could be useful in general. We might want to add backend and run_options to it since they might be needed for some post-processing (eg for adding measurement level metadata for a calibration experiment)

def _postprocess_transpiled_circuits(self, circuits, backend, **run_options): """Additional post-processing of transpiled circuits before running on backend""" pass

qiskit_experiments/randomized_benchmarking/rb_analysis.py

qiskit_experiments/randomized_benchmarking/rb_experiment.py

qiskit_experiments/randomized_benchmarking/rb_utils.py

nkanazawa1989 · 2021-06-18T03:12:54Z

qiskit_experiments/randomized_benchmarking/rb_utils.py

+    @staticmethod
+    def coherence_limit(nQ=2, T1_list=None, T2_list=None, gatelen=0.1):
+        """
+        The error per gate (1-average_gate_fidelity) given by the T1,T2 limit.


Can you write physics behind this? This is quite convenient but I felt some difficulty of using this in journal parpers, because we don't provide any technical reference for this calculation.

Note that this code was copied from the ignis code written by Dave, so you should refer this question to him:
https://github.com/Qiskit/qiskit-ignis/blob/9201ed8f9970c50308601f4dfd895a1b8255e469/qiskit/ignis/verification/randomized_benchmarking/rb_utils.py#L169-#L224

This is not originally my code, but I'll try.

Actually I don't know the original paper of this calculation. Seems like some conventional code used by IBM experimentalists. One thing really close to this form is shown in Appendix G of this paper.

It would be great if you can derive the coherence limit, but this is not mandatory for this PR. Can you write an issue for future extension?

nkanazawa1989 · 2021-06-18T03:29:38Z

qiskit_experiments/randomized_benchmarking/rb_utils.py

+                    found_gates.append(gate)
+                    error_sum += gates_per_clifford[key] * value
+            for gate in found_gates:
+                epg[qubit][gate] = (error_dict[((qubit,), gate)] * epc_1_qubit) / error_sum


It would be better to keep sigma.

Not sure I follow you - sigma instead of what?

Tough EPC has error (sigma=stdev), this information is not used by EPG. Perhaps this is future extension. Can you write an issue for reminder?

Since EPG is calculated from EPC, it should be possible to calculate the error bounds for EPG.

test/test_rb.py

…kend, we pass error dict

nkanazawa1989

Thanks Gadi, the structure looks good. There are some nitpicks and test idea. Let's hear opinion of @chriseclectic about base class change.

qiskit_experiments/randomized_benchmarking/rb_experiment.py

nkanazawa1989 · 2021-06-22T07:44:20Z

qiskit_experiments/randomized_benchmarking/rb_utils.py

+    """A collection of utility functions for computing additional data
+    from randomized benchmarking experiments"""
+
+    @staticmethod


Perhaps you want to decorate with @classmethod? Same for others. (i.e. there is no code creating an instance of RBUtil)

This is what @staticmethod does. In @classmethod the class itself is a parameter (for setting class variables etc.) and I don't need it here.

test/test_rb.py

qiskit_experiments/randomized_benchmarking/rb_experiment.py

* Added the rb_utils file from Ignis as a class * Adding transpiled circuits postprocessing * Computing gate count by qubit * gates_per_clifford was simplified and removed from rb_utils * Getting the error dict from the backend * Computing gates per clifford * EPG for 1 qubit is now working * Refactoring for EPG on 2 qubits * 1 qubit epg is working after refactoring * 2 qubit epg is done but might have bugs * Finalizing epg analysis and removing debug prints * Added documentation * Linting * Small fix * Integrating the curve analysis class with the epg computation in RB * Added epg computation test * Bugfix in n_gate_2q computation * Passing RB analysis data via options * Linting * Linting * Linting * Lower test sensitivity * Adding more lengths to the rb test * Linting * Temporarily disabling test due to nondeterministic behavior on testing machine * Temporarily disabling test due to nondeterministic behavior on testing machine * Gates per clifford is now computed in the analysis section * gate_error_ratio is now an analysis option and instead of passing backend, we pass error dict * Linting * Small fix * Small fixes * Added count_ops test * Added calculate_1q_epg test * Linting * Linting * Moving error_dict computation from experiment to analysis

gadial and others added 4 commits June 1, 2021 16:12

Added the rb_utils file from Ignis as a class

0b14f7f

Adding transpiled circuits postprocessing

ce68e02

Computing gate count by qubit

bac0cff

gates_per_clifford was simplified and removed from rb_utils

7f85816

gadial changed the title ~~Rb utils class~~ [WIP] RB utils class Jun 7, 2021

gadial and others added 10 commits June 8, 2021 15:38

Getting the error dict from the backend

e3fd85e

Computing gates per clifford

517aeb9

EPG for 1 qubit is now working

f84c740

Refactoring for EPG on 2 qubits

79311a4

1 qubit epg is working after refactoring

845a41b

2 qubit epg is done but might have bugs

272159e

Finalizing epg analysis and removing debug prints

1102f82

Added documentation

82bccfb

Linting

ecfb295

Small fix

687ea1b

gadial changed the title ~~[WIP] RB utils class~~ RB utils class Jun 10, 2021

gadial mentioned this pull request Jun 10, 2021

RB Utils file #61

Closed

gadial added 3 commits June 10, 2021 16:13

Merge branch 'main' into rb_utils_class

822b613

Integrating the curve analysis class with the epg computation in RB

15602b2

Added epg computation test

dc8fcf4

nkanazawa1989 reviewed Jun 11, 2021

View reviewed changes

ShellyGarion added this to the Release 1 milestone Jun 14, 2021

gadial and others added 8 commits June 14, 2021 17:34

Merge branch 'main' into rb_utils_class

37d71c3

Merge branch 'main' into rb_utils_class

7a91529

Bugfix in n_gate_2q computation

8181b17

Passing RB analysis data via options

a0caff8

Linting

cd97183

Linting

ed0e09e

Linting

ef0504b

Lower test sensitivity

8a0a274

gadial added 4 commits June 16, 2021 17:35

Adding more lengths to the rb test

c2a99cc

Linting

4f99de1

Merge branch 'main' into rb_utils_class

69b8bc5

Temporarily disabling test due to nondeterministic behavior on testin…

7d7a728

…g machine

gadial requested a review from ShellyGarion June 17, 2021 08:43

ShellyGarion reviewed Jun 17, 2021

View reviewed changes

Temporarily disabling test due to nondeterministic behavior on testin…

f92d02f

…g machine

nkanazawa1989 requested changes Jun 18, 2021

View reviewed changes

gadial and others added 4 commits June 21, 2021 08:45

Merge branch 'main' into rb_utils_class

b4382e8

Gates per clifford is now computed in the analysis section

d963839

gate_error_ratio is now an analysis option and instead of passing bac…

080a5bc

…kend, we pass error dict

Linting

95c05a6

nkanazawa1989 requested changes Jun 22, 2021

View reviewed changes

gadial added 7 commits June 22, 2021 13:38

Small fix

08b616d

Small fixes

1dc0a21

Added count_ops test

6dd2e4c

Added calculate_1q_epg test

61187b7

Merge branch 'main' into rb_utils_class

5f2d8ef

Linting

27b3a5e

Linting

8d7e0ad

chriseclectic reviewed Jun 23, 2021

View reviewed changes

qiskit_experiments/randomized_benchmarking/rb_experiment.py Outdated Show resolved Hide resolved

chriseclectic mentioned this pull request Jun 23, 2021

QV experiment #32

Merged

gadial added 2 commits June 23, 2021 22:01

Moving error_dict computation from experiment to analysis

a2ff6d0

Merge branch 'main' into rb_utils_class

ec4bd25

nkanazawa1989 approved these changes Jun 24, 2021

View reviewed changes

gadial merged commit 5476c68 into qiskit-community:main Jun 24, 2021

RB utils class #86

RB utils class #86

Uh oh!

Conversation

gadial commented Jun 7, 2021

Summary

Details and comments

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!

ShellyGarion left a comment

Choose a reason for hiding this comment

Uh oh!

nkanazawa1989 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!

chriseclectic Jun 23, 2021 • 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!

Uh oh!

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!

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!

nkanazawa1989 Jun 22, 2021 • 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!

Uh oh!

nkanazawa1989 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

chriseclectic Jun 23, 2021 •

edited

Loading

nkanazawa1989 Jun 22, 2021 •

edited

Loading