sandialabs / LoQS / 31952745738

88%

push

github

Multi-qubit, multi-outcome instruments in statevector backend (#110)

* Enable 2Q/4Q parity-check instruments in the statevector backend (#51)

ZBasisOutcomeOperationDictInstrumentRep gains an outcome_qubits field,
decoupled from qubit_labels, so a single classical channel can carry
arbitrary outcome keys (a joint 2Q/4Q parity-check instrument) while
multiple channels stay decomposable per-qubit. npsvstate.py's instrument
handler is generalized on top of a new shared _sample_lazy helper
(factored out of the existing Kraus sampler, verified byte-for-byte
unchanged before adding new logic), removing the old 1-qubit/2-outcome
restriction. PyGSTiNoiseModel gains an instrument_outcome_qubits
parameter to declare which classical register a joint instrument's
outcome belongs to.

- instrumentreps.py: outcome_qubits field, key-shape validation,
  length-1 tuple-key normalization (fixes a latent bare-int/tuple key
  mismatch), and a with_qubit_labels override that only retargets
  outcome_qubits when it still matches qubit_labels.
- npsvstate.py: _sample_lazy extracted from _apply_kraus_lazy;
  instrument handler rewritten on top of it, dropping the old
  NotImplementedError/{0,1}-only restriction.
- pygstimodel.py: instrument_outcome_qubits constructor parameter;
  _get_instrument_rep detects joint vs. decomposable outcome shape and
  accepts non-numeric member labels (e.g. 'even'/'odd').
- conversion.py: no functional change, regression-tested against the
  new field.
- qsimstate.py: unchanged, still 1-qubit-only; deferred to a
  follow-up branch.

Tests cover decomposable and joint multiqubit instruments, a 3-outcome
joint channel, include_outcome suppression, lazy-stopping, and an
end-to-end PyGSTiNoiseModel -> get_reps -> SVState case. Full
tests/backends+core+internal suite passes (746 passed, 2 skipped, 2
xfailed); mypy error count unchanged from main (87).

* Add parity check tutorial

* Tutorial updates and CHANGELOG for #110

* Fix Imrz missed in merge

1345 of 1479 branches covered (90.94%)

Branch coverage included in aggregate %.

7399 of 8363 relevant lines covered (88.47%)

0.88 hits per line