Add support for Python 3.13

[edoaltamura]

This update introduces support for Python 3.13, scheduled for final release on 7 October 2024.

Closes #825
Closes OkuyanBoga#45

[woodsp-ibm]

FYI: Qiskit 1.3.0 milestone, due by November 21, 2024

• Add official support for Python 3.13 Qiskit/qiskit#13309

[edoaltamura]

Un-drafting this PR ahead of the RC expected release date of Qiskit 1.3 (7 Nov 2024). The due date for the official release remains 21 Nov, which is when we expect to sync Python 3.13 support for Qiskit ML too.

[edoaltamura]

This PR has to be put on hold until #863 is resolved.

[edoaltamura]

Thanks very much @woodsp-ibm! I added the __eq__ functions to the optimizers to reflect those in Algos.

[woodsp-ibm]

I see my eq change failed with mypy - as there are so many mypy issues now in qiskit algorithms I had disabled it there as an interim solution so I had not seen that. Presumably if the object instance type is checked why would it not return False if its another type in this case rather than NotImplemented as the suggestion.

[edoaltamura]

Note that mypy requires this construct def __eq__(self, other: object) -> bool:, so I changed the methods to

# Under Python 3.13 the auto-generated equal fails with an error around
    # using numpy all or any. See https://github.com/qiskit-community/qiskit-algorithms/pull/225
    # for further information. Hence, this custom function was added. The __eq__
    # method is supposed to accept any object. If you update the version of
    # mypy you're using, it'll print out a note recommending this code structure.
    def __eq__(self, other: object) -> bool:
        if not isinstance(other, OptimizerState):
            # If we return NotImplemented, Python will automatically try
            # running other.__eq__(self), in case 'other' knows what to do with
            # Person objects.
            return NotImplemented

        return (
            (
                self.x == other.x
                if isinstance(self.x, float)
                else (self.x.shape == other.x.shape and (self.x == other.x).all())
            )
            and self.fun == other.fun
            and self.jac == other.jac
            and self.nfev == other.nfev
            and self.njev == other.njev
            and self.nit == other.nit
        )

and

# See parent class for a comment on having a custom equals. I needed this
    # too as it does not appear to use super by default and without this failed
    # the exact same way. Note it does not include learning rate as that field
    # is not included in the compare as pre the field decorator. The __eq__
    # method is supposed to accept any object. If you update the version of
    # mypy you're using, it'll print out a note recommending this code
    # structure.
    def __eq__(self, other: object) -> bool:
        if not isinstance(other, GradientDescentState):
            # If we return NotImplemented, Python will automatically try
            # running other.__eq__(self), in case 'other' knows what to do with
            # Person objects.
            return NotImplemented

        return super().__eq__(other) and self.stepsize == other.stepsize

You might find these useful for Algorithms too.

[woodsp-ibm]

Thanks yes. I had already made a mental note to update algs with whatever worked here.

[woodsp-ibm]

I am not sure why mypy is complaining now since that ternary was to do something different for float type such that it does not do the shape check etc

[edoaltamura]

Yeah I'm confused by mypy. It's basically forcing to cast many function declarations, and sometimes even confuses int for str.

mypy qiskit_machine_learning test tools
qiskit_machine_learning/state_fidelities/base_state_fidelity.py:312: error: Incompatible return value type (got "str | list[str] | list[list[str]] | list[list[list[Any]]]", expected "Sequence[float]")  [return-value]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("float" and "list[float]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("float" and "list[list[float]]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("float" and "list[list[list[Any]]]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: error: No overload variant of "__add__" of "list" matches argument type "float"  [operator]
qiskit_machine_learning/gradients/utils.py:107: note: Possible overload variants:
qiskit_machine_learning/gradients/utils.py:107: note:     def __add__(self, list[float], /) -> list[float]
qiskit_machine_learning/gradients/utils.py:107: note:     def [_S] __add__(self, list[_S], /) -> list[_S | float]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("list[float]" and "list[list[float]]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("list[float]" and "list[list[list[Any]]]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: note:     def __add__(self, list[list[float]], /) -> list[list[float]]
qiskit_machine_learning/gradients/utils.py:107: note:     def [_S] __add__(self, list[_S], /) -> list[_S | list[float]]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("list[list[float]]" and "list[float]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("list[list[float]]" and "list[list[list[Any]]]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: note:     def __add__(self, list[list[list[Any]]], /) -> list[list[list[Any]]]
qiskit_machine_learning/gradients/utils.py:107: note:     def [_S] __add__(self, list[_S], /) -> list[_S | list[list[Any]]]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("list[list[list[Any]]]" and "list[float]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: error: Unsupported operand types for + ("list[list[list[Any]]]" and "list[list[float]]")  [operator]
qiskit_machine_learning/gradients/utils.py:107: note: Both left and right operands are unions
qiskit_machine_learning/gradients/utils.py:107: error: Incompatible return value type (got "float | Any | list[float] | list[list[float]] | list[list[list[Any]]]", expected "list[ndarray[Any, Any]]")  [return-value]
qiskit_machine_learning/optimizers/steppable_optimizer.py:92: error: Item "float" of "float | ndarray[Any, Any]" has no attribute "shape"  [union-attr]
qiskit_machine_learning/optimizers/aqgd.py:170: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[Any]] | ndarray[tuple[int, ...], dtype[float64]]", variable has type "list[ndarray[tuple[int, ...], dtype[Any]] | ndarray[tuple[int, ...], dtype[float64]]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:187: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[Any]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:189: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[Any]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:190: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[Any]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:216: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[float64]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:217: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[float64]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:219: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[float64]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
qiskit_machine_learning/optimizers/adam_amsgrad.py:234: error: Incompatible types in assignment (expression has type "ndarray[tuple[int, ...], dtype[Any]]", variable has type "ndarray[tuple[int], dtype[float64]]")  [assignment]
Found 21 errors in 5 files (checked 185 source files)

[woodsp-ibm]

Is that mypy output from your local machine? Otherwise, as far as I can see, there is just the one issue its flagging in CI here - maybe locally you are not using the same versions as CI is.

[woodsp-ibm]

For the mypy shape failure it seems my ternary expression had issues on closer inspection. I.e the first part said if both are float but it could then end up in the else where just one was a float, which evidently wont have shape. So it became a bit more complex and I updated the algs with the other mypy issue as well as fixing the shape aspect here qiskit-community/qiskit-algorithms#226 (it was failing with the mypy issue the same way locally but now passes locally with mypy for me though as I mentioned its disabled at present in algs since thats using a newer version of numpy - locally I still have a 2.0.2)

[edoaltamura]

It worked with the fix in Algorithms!

[coveralls]

Pull Request Test Coverage Report for Build 14001256542

Details

• 6 of 8 (75.0%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.03%) to 90.828%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
qiskit_machine_learning/optimizers/gradient_descent.py	3	4	75.0%
qiskit_machine_learning/optimizers/steppable_optimizer.py	3	4	75.0%

Totals
Change from base Build 13973547049:	-0.03%
Covered Lines:	4486
Relevant Lines:	4939

---

💛 - Coveralls

[woodsp-ibm]

Just needs the branch rules updated then to remove the required 3.12 jobs no longer run and make the 3.13 ones, that superseded them, required,

[edoaltamura]

Thanks for approving the PR @woodsp-ibm! If I remember correctly, I cannot change the rules myself because only qiskit community members have those privileges. If this is still the case, could you please help with that?

[woodsp-ibm]

@edoaltamura Ok I updated the rules for the main branch here. I will note in looking @smens has admin rights here in this repo and therefore should be able to do such a change too,