Benchmarking of hardware acceleration of pyhf
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: run.py [OPTIONS]
Automatic process of taking pyhf computation.
Usage:
$ python run.py -c [-b] [-p] [-u] [-m] [-n] [-mm]
Examples:
$ python run.py -c mle -b numpy -u https://www.hepdata.net/record/resource/1267798?view=true -m [750,100]
$ python run.py -c mle -u https://www.hepdata.net/record/resource/1267798?view=true -m [750,100]
$ python run.py -c mle -b numpy -p 1Lbb-likelihoods-hepdata -m [750,100]
$ python run.py -c interpolation -b jax -n 0 -mm fast
$ python run.py -c interpolation -b numpy -n 0 -mm slow
More information:
https://github.com/pyhf/pyhf-benchmark
Options:
-c, --computation TEXT Type of computation.
-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 Interpolation type.
-mm, --mode TEXT Interpolation mode.
--help Show this message and exit.
pyhf-benchmark is openly developed by Bo Zheng and the pyhf dev team.
Please check the contribution statistics for a list of contributors.
Bo Zheng was awarded an IRIS-HEP Fellowship for this work.