https://github.com/pennylaneai/pulse-benchmarks
Some basic benchmarks for `pennylane.pulse`
https://github.com/pennylaneai/pulse-benchmarks
Last synced: 10 months ago
JSON representation
Some basic benchmarks for `pennylane.pulse`
- Host: GitHub
- URL: https://github.com/pennylaneai/pulse-benchmarks
- Owner: PennyLaneAI
- Created: 2023-09-15T09:09:53.000Z (almost 3 years ago)
- Default Branch: main
- Last Pushed: 2023-12-07T15:30:34.000Z (over 2 years ago)
- Last Synced: 2024-12-23T00:05:20.868Z (over 1 year ago)
- Language: Jupyter Notebook
- Size: 2.64 MB
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# pulse-benchmarks
Some basic benchmarks for `pennylane.pulse` for public display.
We run the basic benchmark of evolving a state according to
$$ H(t) = \sum_q \omega_q Z_q + \sum_q g_q Y_q \otimes Y_{q+1} + \sum_q f_q(t) X_q \otimes X_{q+1} $$
and then computing the expectation value of
$$ H_\text{obj} = \sum_q c^X_q X_q \otimes X_{q+1} + c^Y_q Y_q \otimes Y_{q+1}. $$
As time-dependent envelopes $f_q(t)$, we use a (modulated) Gaussian in resonance with the qubit frequencies $\omega_q = 5$ GHz.
The main benchmark is done in [Forward-pass_PL_QUTIP_QISKIT.ipynb](Forward-pass_PL_QUTIP_QISKIT.ipynb). We ran this on an Macbook Air 15 with an Apple M2 processor. We get qualitatively identical results using an Intel i7-1260P in an LG gram 16 (2022). Please open an issue in case you experience significant deviations from this picture
We also perform the same benchmark on a NVIDIA TITAN Xp GPU, see [Forward-pass_PL_hardware.ipynb](Forward-pass_PL_hardware.ipynb).
## Disclaimer
We compare to third party repositories offering similar functionality. Benchmarks never capture the full capability of either repository and are always specific to the choice of benchmark problem. Further, we cannot guarantee using the latest and best standards for any third party repository that we compare to. We invite experts to submit suggestions for better performance by opening Issues and/or Pull Requests.