- Moved aqc MPS content into aqc/mps sub-module.

[ACE07-Sev]

• Added aqc/sweep module, which provides StateSweepPass and UnitarySweepPass for performing approximate synthesis for both state and unitary using tensor network sweeping.
• Added unit tests for aqc/sweep.
• Added .mypy_cache/ to .gitignore file.

closes #368 .

[ACE07-Sev]

@bachase Here is the sweeping we talked about. I added state to it as well.

[ACE07-Sev]

Oops, forgot doctest.

[ACE07-Sev]

Once I finish qmprs PR to incorporate bond 2 compression optimization, you can then use that MPS version + SweepStateApproximator to get the performance improvement I mentioned for MPS.

[ACE07-Sev]

@bachase Have you had a chance to review the PR?

[bachase]

Sorry @ACE07-Sev I had lost track of this one. Taking a look now

[Copilot]

Pull request overview

This PR reorganizes the AQC transpiler by moving existing MPS-based approximate compilation into an aqc/mps submodule and introducing a new aqc/sweep submodule that adds sweeping-based approximate synthesis passes for both states and unitaries (with accompanying tests and docs updates).

Changes:

• Moved MPS AQC implementation into ucc/transpilers/aqc/mps and updated imports/docs accordingly.
• Added ucc/transpilers/aqc/sweep with StateSweepPass / UnitarySweepPass and approximate compilation helpers.
• Added tests for the new sweep passes and updated .gitignore for .mypy_cache/.

Reviewed changes

Copilot reviewed 15 out of 17 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
ucc/transpilers/aqc/utils.py	New shared helper for memory checks (moved from MPS utils).
ucc/transpilers/aqc/sweep/__init__.py	Public sweep API and compile helpers (approx_state_compile, approx_unitary_compile).
ucc/transpilers/aqc/sweep/sweep_pass.py	New Qiskit TransformationPass wrappers for sweep compilation.
ucc/transpilers/aqc/sweep/approx_state.py	State sweeping approximator implementation (tensor-network based).
ucc/transpilers/aqc/sweep/approx_unitary.py	Unitary sweeping approximator implementation + fidelity helper.
ucc/transpilers/aqc/sweep/ansatzes.py	Ansatz generators (staircase / brickwall) for sweep methods.
ucc/transpilers/aqc/sweep/typing.py	Type aliases for unitary layer/block structures.
ucc/transpilers/aqc/mps/utils.py	Removes has_enough_memory after moving to shared AQC utils.
ucc/transpilers/aqc/mps/__init__.py	New MPS submodule public API and approx_compile.
ucc/transpilers/aqc/mps/mps_pass.py	Imports approx_compile lazily to avoid module import cycles.
ucc/transpilers/aqc/mps/mps_sequential.py	Updates slope util import and fixes fidelity computations to use squared magnitude.
ucc/transpilers/aqc/mps/qmprs_compiler.py	Updates slope util import and adds typing ignore for optional dependency.
ucc/transpilers/aqc/mps/README.md	New documentation describing the MPS approach and qmprs option.
ucc/transpilers/aqc/__init__.py	Re-exports passes from the new submodules.
ucc/tests/test_compile.py	Adds sweep-pass tests and updates imports.
docs/source/user_guide.rst	Updates usage example import path for MPSPass.
.gitignore	Ignores .mypy_cache/.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[bachase]

Left some initial comments on just the tests. Copilot had some good ones too. I'll do another deeper review soon.

[bachase]

Are random states over this distribution equally interesting/exercising the protocol well?

[ACE07-Sev]

I believe it's the general random state distribution. I can test on other distributions as well.

[ACE07-Sev]

Saw the review. I will go through them today.

[ACE07-Sev]

Sorry for the delay. I am going through it now.

[ACE07-Sev]

@bachase I amended for the review comments. One major thing that needs to be done sometime is finding a good rule for defining the number of layers. Its optimal value very much depends on the state/unitary structure and size. There's a dynamic way I can try. Basically:

- Add layer
- Perform sweeps until plateau
- If plateau is lower than desired lowerbound then repeat, else stop

This will remove the hardcoded values, but in certain cases may take too long to reach the desired fidelity. Although, we can introduce hard stops like if the ansatz generated is more expensive than the original circuit.

[ACE07-Sev]

Actually, do you want me to do that now before I push? The dynamic approach?

[bachase]

Thanks for investigating @ACE07-Sev . How "big" of a change/exploration is the dynamic approach? If you think its worth doing now, go for it, but ok if you think a followup is better.