QuOptuna
Quantum Classical Optuna + PennyLane
One search. Quantum & classical.
Fairness-aware, explainable hyperparameter tuning across both model families.
# one command, no repo, no Node
$ uvx quoptuna optimize --uci-id 53 --trials 30
trial 12/30 · IQPKernelf1=0.91
trial 24/30 · SVCf1=0.94
✓ best: SVC · macro-F1 0.947 · study saved
Open source · MIT
Star it if it saved you a trial.
Stars help other researchers find QuOptuna and steer what we build next.
—
GitHub stars
—
PyPI / month
Capabilities
Everything a real run needs.
6-step wizard
Dataset → Report. State persists across restarts.
Samplers & pruners
TPE, random, grid — with ASHA / Hyperband early stopping.
Fairness-aware
Constrained or multi-objective on disparity metrics.
Explainable
SHAP, metric curves, confusion matrices on demand.
AI reports
Governance-ready summaries via OpenAI, Gemini, or Anthropic.
Multiclass & durable
One-vs-rest macro-F1, SQLite runs that survive crashes.
Model catalog
Two search spaces, one objective.
Quantum · PennyLane
9- CircuitCentric
- DataReuploading
- DressedQuantumCircuit
- ProjectedQuantumKernel
- QuantumKitchenSinks
- QuantumMetricLearner
- TreeTensor
- IQPKernel
- Separable
one objective
Classical · scikit-learn
4- SVC
- SVC (linear)
- MLPClassifier
- Perceptron
End to end
From a CSV to a report.
1
Dataset
CSV or UCI
2
Features
Pick & encode
3
Configure
Models, sampler
4
Optimize
Live trials
5
Analyze
SHAP & metrics
6
Report
AI summary
Boot the whole app with one command.
uvx quoptuna serves the UI and API on a single port — no Node, no checkout.