Fix a bug of missing pulse library entry in PulseQobj parsing #11397

nkanazawa1989 · 2023-12-11T17:29:29Z

Summary

When a user tries to get calibration data from Target, it may fail because of partly missing calibration data, e.g.

backend.target["some_gate"][(0, 1)].calibration

This may occur when the backend doesn't support waveform payload, while Qiskit doesn't have definition of the corresponding parametric pulse that the backend uses.

Details and comments

This PR adds exception handling for this case. Namely, None value is returned instead of raising an error.

qiskit-bot · 2023-12-11T17:29:33Z

One or more of the the following people are requested to review this:

@Qiskit/terra-core
@nkanazawa1989

coveralls · 2023-12-11T18:07:08Z

Pull Request Test Coverage Report for Build 7532191522

Warning: This coverage report may be inaccurate.

We've detected an issue with your CI configuration that might affect the accuracy of this pull request's coverage report.
To ensure accuracy in future PRs, please see these guidelines.
A quick fix for this PR: rebase it; your next report should be accurate.

0 of 0 changed or added relevant lines in 0 files are covered.
No unchanged relevant lines lost coverage.
Overall coverage increased (+1.8%) to 89.354%

Totals
Change from base Build 7463210826:	1.8%
Covered Lines:	59498
Relevant Lines:	66587

💛 - Coveralls

wshanks

The code changes look okay here, but I would a little more explanation of the issue. I don't know all the details of the qobj code, but I am guessing only a few people at most know it better than me, so if I am having trouble understanding what the issue is here probably most other people will as well.

From the release note, I understand that the problem occurs when the backend does not support a waveform payload while Qiskit does not have a definition for the parametric pulse shape reported. So is the situation that the defaults json includes a parametric_pulse entry with a name that is not in the ParametricPulseShapes Enum? So when the converter tries to convert a parametric pulse, it hits an AttributeError here:

qiskit/qiskit/qobj/converters/pulse_instruction.py

Line 920 in b8d21ae

    
           pulse = ParametricPulseShapes.to_type(instruction.pulse_shape)(**params, name=pulse_name)

because to_type uses getattr:

qiskit/qiskit/qobj/converters/pulse_instruction.py

Lines 80 to 89 in b8d21ae

    
               def to_type(cls, name: str) -> library.SymbolicPulse: 
        
                   """Get symbolic pulse class from the name. 
        
                   Args: 
        
                       name: Qobj name of the pulse. 
        
                   Returns: 
        
                       Corresponding class. 
        
                   """ 
        
                   return getattr(library, cls[name].value)

and that causes the converter to fall back to _convert_generic here:

qiskit/qiskit/qobj/converters/pulse_instruction.py

Lines 662 to 665 in b8d21ae

    
           try: 
        
               method = getattr(self, f"_convert_{instruction.name}") 
        
           except AttributeError: 
        
               method = self._convert_generic

If that is the case, I don't think the part about supporting a waveform payload matters? You wouldn't expect a pulse to be in the waveform pulse library if it was reported as a parametric pulse.

wshanks · 2024-01-10T17:20:08Z

qiskit/qobj/converters/pulse_instruction.py

+            if qubits := getattr(instruction, "qubits", None):
+                msg = f"qubits {qubits}"
+            else:
+                msg = f"channel {instruction.ch}"


ch does not have to be set as an attribute of PulseQobjInstruction either. Should we just hope that either qubits or ch is set? Every instruction should have a ch set except acquire which should have qubits set so I think for a valid qobj that is reasonable (it is weird that this block assumed qubits would always be defined here rather than ch).

Yeah both ch and qubits are optional as you say. For now having either ch and qubits is valid assumption.

wshanks · 2024-01-10T18:00:28Z

releasenotes/notes/fix-missing-pulse-lib-c370f5b9393d0df6.yaml

+    This occurs when the backend doesn't support pulse waveform payload, while
+    Qiskit doesn't have corresponding parametric pulse definition.
+    In this situation, Qiskit pulse object cannot be built, resulting in the failure in
+    parsing PulseQobj and the user encounters a meaningless error.


What is the situation where the error is meaningless? This changes the code so that the case where a calibration schedule was unloadable behaves like no calibration was reported by the backend rather than raising an exception. If the user wanted to use the calibration, the user's code will still have a problem. Maybe there is some code that eagerly handles calibrations and can work okay when they are not available?

Qobj is parsed when they are called for the first time (i.e. when a user call .calibration of target instruction) and the meaningless error raised when the user called calibration in their program (non-pulse user will never see this). It tries to raise a QiskitError but the error message required .qubits that the play instruction payload cannot provide, and this resulted in AttributeError instead. It was very hard to understand what was happening.

I expect the user implements the code that checks if calibration is available on their side.

wshanks · 2024-01-15T15:44:30Z

qiskit/pulse/calibration_entries.py

+        except QiskitError:
+            # When the play waveform data is missing in pulse_lib we cannot build schedule.
+            # Instead of raising an error, get_schedule should return None.
+            self._definition = IncompletePulseQobj


Could we emit a warning (UserWarning?) here with the message from the QiskitError? Here we are treating a calibration loading error as equivalent to no calibration being provided and the user might want to know when there is a loading error.

Fair enough. Added an error and test in bf723eb

releasenotes/notes/fix-missing-pulse-lib-c370f5b9393d0df6.yaml

Co-authored-by: Will Shanks <wshaos@posteo.net>

wshanks

Looks good!

ElePT

Approving to unblock the merge, thanks a lot!

* Fix missing pulse lib bug * Update releasenotes/notes/fix-missing-pulse-lib-c370f5b9393d0df6.yaml Co-authored-by: Will Shanks <wshaos@posteo.net> * Add user warning --------- Co-authored-by: Will Shanks <wshaos@posteo.net> (cherry picked from commit a9e9e95) # Conflicts: # qiskit/pulse/calibration_entries.py

#11397) (#11573) * Fix a bug of missing pulse library entry in PulseQobj parsing (#11397) * Fix missing pulse lib bug * Update releasenotes/notes/fix-missing-pulse-lib-c370f5b9393d0df6.yaml Co-authored-by: Will Shanks <wshaos@posteo.net> * Add user warning --------- Co-authored-by: Will Shanks <wshaos@posteo.net> (cherry picked from commit a9e9e95) # Conflicts: # qiskit/pulse/calibration_entries.py * Fix merge conflict --------- Co-authored-by: Naoki Kanazawa <nkanazawa1989@gmail.com> Co-authored-by: Elena Peña Tapia <57907331+ElePT@users.noreply.github.com>

nkanazawa1989 added Changelog: Bugfix Include in the "Fixed" section of the changelog mod: pulse Related to the Pulse module labels Dec 11, 2023

nkanazawa1989 requested review from eggerdj, wshanks and a team as code owners December 11, 2023 17:29

nkanazawa1989 force-pushed the fix/missing_pulse_lib branch from 8956e86 to fd71bf8 Compare December 11, 2023 17:42

mtreinish added this to the 0.45.2 milestone Dec 11, 2023

mtreinish added the stable backport potential The bug might be minimal and/or import enough to be port to stable label Dec 11, 2023

Fix missing pulse lib bug

b939e0c

nkanazawa1989 force-pushed the fix/missing_pulse_lib branch from fd71bf8 to b939e0c Compare January 9, 2024 18:14

wshanks self-assigned this Jan 10, 2024

wshanks reviewed Jan 10, 2024

View reviewed changes

wshanks reviewed Jan 15, 2024

View reviewed changes

nkanazawa1989 and others added 2 commits January 16, 2024 01:32

Update releasenotes/notes/fix-missing-pulse-lib-c370f5b9393d0df6.yaml

4479e2b

Co-authored-by: Will Shanks <wshaos@posteo.net>

Add user warning

bf723eb

wshanks approved these changes Jan 16, 2024

View reviewed changes

ElePT approved these changes Jan 16, 2024

View reviewed changes

ElePT added this pull request to the merge queue Jan 16, 2024

Merged via the queue into Qiskit:main with commit a9e9e95 Jan 16, 2024
13 checks passed

mergify bot mentioned this pull request Jan 16, 2024

Fix a bug of missing pulse library entry in PulseQobj parsing (backport #11397) #11573

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix a bug of missing pulse library entry in PulseQobj parsing #11397

Fix a bug of missing pulse library entry in PulseQobj parsing #11397

nkanazawa1989 commented Dec 11, 2023

qiskit-bot commented Dec 11, 2023

coveralls commented Dec 11, 2023 •

edited

Loading

wshanks left a comment •

edited

Loading

wshanks Jan 10, 2024

nkanazawa1989 Jan 15, 2024 •

edited

Loading

wshanks Jan 10, 2024

nkanazawa1989 Jan 15, 2024

wshanks Jan 15, 2024

nkanazawa1989 Jan 15, 2024

wshanks left a comment

ElePT left a comment

	def to_type(cls, name: str) -> library.SymbolicPulse:
	"""Get symbolic pulse class from the name.

	Args:
	name: Qobj name of the pulse.

	Returns:
	Corresponding class.
	"""
	return getattr(library, cls[name].value)

	try:
	method = getattr(self, f"_convert_{instruction.name}")
	except AttributeError:
	method = self._convert_generic

Fix a bug of missing pulse library entry in PulseQobj parsing #11397

Fix a bug of missing pulse library entry in PulseQobj parsing #11397

Conversation

nkanazawa1989 commented Dec 11, 2023

Summary

Details and comments

qiskit-bot commented Dec 11, 2023

coveralls commented Dec 11, 2023 • edited Loading

Pull Request Test Coverage Report for Build 7532191522

Warning: This coverage report may be inaccurate.

💛 - Coveralls

wshanks left a comment • edited Loading

Choose a reason for hiding this comment

wshanks Jan 10, 2024

Choose a reason for hiding this comment

nkanazawa1989 Jan 15, 2024 • edited Loading

Choose a reason for hiding this comment

wshanks Jan 10, 2024

Choose a reason for hiding this comment

nkanazawa1989 Jan 15, 2024

Choose a reason for hiding this comment

wshanks Jan 15, 2024

Choose a reason for hiding this comment

nkanazawa1989 Jan 15, 2024

Choose a reason for hiding this comment

wshanks left a comment

Choose a reason for hiding this comment

ElePT left a comment

Choose a reason for hiding this comment

coveralls commented Dec 11, 2023 •

edited

Loading

wshanks left a comment •

edited

Loading

nkanazawa1989 Jan 15, 2024 •

edited

Loading