Setup#
Installation#
Using pip
#
To install specific benchmark dependancies, you can specify any or all of the following:
From source#
To install specific benchmark dependancies, you can specify any or all of the following:
CLI#
mfpbench
provides some helpful commands for downloading data as well as generating priors
for benchmarks.
usage: python -m mfpbench [-h] {install,download,priors} ...
positional arguments:
{install,download,priors}
install python -m mfpbench install --help
Install requirements for use with mfpbench.
-------------------------------------------
download python -m mfpbench download --help
Download datasets for use with mfpbench.
----------------------------------------
priors python -m mfpbench priors --help
Generate priors for use with mfpbench.
--------------------------------------
options:
-h, --help show this help message and exit
Download#
Some benchmark require raw files to work, you can download these using the
python -m mfpbench download
command.
You can specify the location of the root data directory with --data-dir
.
We recommend leaving this to the default to make working with benchmarks easier.
root: /home/runner/work/mf-prior-bench/mf-prior-bench/data
----------------------------------------------------------
[x] yahpo python -m mfpbench download --benchmark yahpo
[x] jahs python -m mfpbench download --benchmark jahs
[x] pd1 python -m mfpbench download --benchmark pd1
[x] lcbench-tabular python -m mfpbench download --benchmark lcbench-tabular
usage: python -m mfpbench download [-h] [--force] [--status] [--list]
[--benchmark {yahpo,jahs,pd1,lcbench-tabular}]
[--data-dir DATA_DIR]
options:
-h, --help show this help message and exit
--force Force download and remove existing data
--status Print out the status of benchmarks
--list Print out the available benchmarks data sources
--benchmark {yahpo,jahs,pd1,lcbench-tabular}
The benchmark to download.
--data-dir DATA_DIR Where to save the data, defaults to './data'
Install#
As there will be conflicting dependancies between different benchmarks, we added some requirements
files to specifically run these benchmarks. You will still have to setup an environment however you
would do so but you can quickly install the required dependancies using the python -m mfpbench install
subcommand.
=====
yahpo
=====
path: /opt/hostedtoolcache/Python/3.11.6/x64/lib/python3.11/site-packages/requirements/yahpo.txt
exec: /opt/hostedtoolcache/Python/3.11.6/x64/bin/python
cmd: python -m pip install -r /opt/hostedtoolcache/Python/3.11.6/x64/lib/python3.11/site-packages/requirements/yahpo.txt
------------------------------------------------------------------------------------------------------------------------
# /opt/hostedtoolcache/Python/3.11.6/x64/lib/python3.11/site-packages/requirements/yahpo.txt
tqdm
usage: python -m mfpbench install [-h] [--list] [--view]
[--benchmark {yahpo,jahs,pd1,lcbench-tabular}]
[--requirements REQUIREMENTS]
options:
-h, --help show this help message and exit
--list Print out the available benchmarks data sources
--view View the default requirements for a benchmark
--benchmark {yahpo,jahs,pd1,lcbench-tabular}
The benchmark to setup.
--requirements REQUIREMENTS
The requirements into the currently active
environment. If not specified, will use the default
requirements. If string, the string will be used as
the path to the requirements file.
Generating Priors#
TODO