CI tests failing on main

[edoaltamura]

Environment

• Qiskit Machine Learning version: 0.8.0
• Qiskit Aer version: qiskit_aer-0.15.1 (stable)
• Qiskit version: 1.3.0

What is happening?

The nightly CI tests on main fail after the Qiskit 1.3.0 release, with error triggered in Aer:

File "C:\hostedtoolcache\windows\Python\3.12.7\x64\Lib\site-packages\qiskit_aer\backends\aerbackend.py", line 478, in _execute_circuits_job
    raise AerError("circuits have parameters but parameter_binds is not specified.")

Occasionally, this behaviour also emits qiskit_machine_learning.exceptions.AlgorithmError: 'Estimator job failed.'.an

How can we reproduce the issue?

Examples of this behaviour can be seen in https://github.com/qiskit-community/qiskit-machine-learning/actions/runs/12111491107at

What should happen?

No error should occur as a consequence of the new release updates; instead, a warning should be emitted in case of upcoming breaking changes.

Any suggestions?

• For version 0.8.0 we can either fix this and bind the parameters to circuits explicitly, label it as stable backport potential and make Qiskit 1.3.0 a requirement, or do this for 0.8.1.
• Fix it internally and keep the QML UI unchanged until later versions, e.g. 0.9.0.

[edoaltamura]

This issue appears to be related to Qiskit/qiskit#13504. For now, we should backport #865 to the 0.8 stable branch, so that:

• [Docs] Fix TOCs and update QNN derived primitives (backport #862) #864 can be backported
• Updated docs can be deployed from stable 0.8
• Users downloading from Pypi don't experience the potential bug with the new Qiskit 1.3 release

#866 could also be addressed at this stage. What are your thoughts on this strategy @OkuyanBoga @woodsp-ibm?