pyhf-benchmark

Benchmarking of hardware acceleration of pyhf

Environment

For the time being, until a library can be created, use the requirements.txt to also serve setup duty in your virtual environment in addition to providing a reproducible benchmarking environment.

(pyhf-benchmark) $ python -m pip install -r requirements.txt

Usage

$ pyhf-benchmark run --help
Usage: pyhf-benchmark run [OPTIONS]

    Automatic process of taking pyhf computation.

    Usage:

    $ pyhf-benchmark run -c [-b] [-p] [-u] [-m] [-n] [-mm]

  Examples:

    $ pyhf-benchmark run -c mle -b numpy -u https://www.hepdata.net/record/resource/1267798?view=true -m [750,100]
    $ pyhf-benchmark run -c mle -u https://www.hepdata.net/record/resource/1267798?view=true -m [750,100]
    $ pyhf-benchmark run -c mle -b numpy -p 1Lbb-likelihoods-hepdata -m [750,100]
    $ pyhf-benchmark run -c interpolation -b jax -n 0 -mm fast
    $ pyhf-benchmark run -c interpolation -b numpy -n 0 -mm slow

  More information:

    https://github.com/pyhf/pyhf-benchmark



Options:
  -c, -computation TEXT   Type of computation  [required]
  -b, --backend TEXT      Name of the pyhf backend to run with.
  -p, --path TEXT         Local path of workspace.
  -u, --url TEXT          Online data link.
  -m, --model-point TEXT  Model point.
  -n, --number TEXT       Number.
  -mm, --mode TEXT        Mode.
  -h, --help              Show this message and exit.

Authors

pyhf-benchmark is openly developed by Bo Zheng and the pyhf dev team.

Please check the contribution statistics for a list of contributors.

Acknowledgements

Bo Zheng was awarded an IRIS-HEP Fellowship for this work.