We present Quarry, an assistant tool for providing fast quantum circuit fidelity estimations across a variety of computing platforms. Quarry processes these estimated values and provides researchers with recommendations for what platforms their particular circuit would perform best on. Our design allows Quarry to use a variety of methods for determining these recommendations. A more detailed explanation and evaluation of the system is available in our final report.
env.yml contains Conda dependencies. Run:
conda env update --file env.yml
to make sure python environment is correctly configured.
General usage form:
python ./src/Est.py file mode
To display help:
python ./src/Est.py -h
Command line arguments
| Argument | Description |
|---|---|
| file | Path to qasm file |
| mode | Query mode |
| -n | Number of platforms to query |
Main file used to query about a qasm file.
Run to generate a data set containing Qiskit backend and circuit fidelity information.
Misc. functions for use in other files.
Methods for calculating fidelity metrics.
-
Is ESP correlated with other metrics?
- Spearman's rank correlation coefficient
- Are ML models actually worth it compared to ESP?
- Use OpenMP to multithread computation.
-
Focus on developing model for swaps
- Compare performance against compiler.
- Train on heavily optimized circuits, save computation estimating in the future
-
Design model for predicting fitness instead of other metrics?
- Could benefit prototype implementation. Only need to train 1 model instead of all the component metric models.
-
Timeline visualization of fitness
-
Heatmap of latency for methods (QC size, QM size) -> Latency
-
Heatmaps of (Metric1, Metric2) -> ESP ?