Add sklearn.show_versions() similar to pandas.show_versions (with numpy blas binding info)

Some numeric issues are related to the specific blas that numpy is using. I’m wondering if it makes sense to add the relevant system_info invocations to the template to make it easier for people to report.

[rth]

@lesteve suggested we could add something like pandas.show_versions() that would print all the relevant information for debugging. For instance, on my laptop, I get,

>>> pd.show_versions()
INSTALLED VERSIONS
------------------
commit: None
python: 3.6.5.final.0
python-bits: 64
OS: Linux
OS-release: 4.11.6-gentoo
machine: x86_64
processor: Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: en_GB.UTF-8

pandas: 0.23.0
pytest: 3.5.1
pip: 10.0.1
setuptools: 39.1.0
Cython: 0.27.3
numpy: 1.14.3
scipy: 1.1.0
pyarrow: None
xarray: None
IPython: 6.4.0
sphinx: None
patsy: None
dateutil: 2.7.3
pytz: 2018.4
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.2.2
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

we certainly don’t care about all the dependencies that pandas might, but I agree that for scikit-learn, having e.g.

• BLAS information
• whether “conda” is in path
• with https://github.com/scikit-learn/scikit-learn/pull/11166 whether the bundled or unbundled joblib is used

would be definitely useful for debuging. It’s more practical to have small function for those, than a copy passable snippet (particularly if it becomes more complex).

Tagging this for v0.20 as this would be fairly easy to do, and help with the maintenance after the release…

[aboucaud]

Good point we can modify only the rst doc for now and delay the changes in .md until the release.

Will do.

I would be in favour of making it accessible at from the root namespace so that you can do from sklearn import show_versions

Agreed. I would have suggested to add from sklearn.utils import show_versions in sklearn.__init__.py instead of putting the file there. Does that sound ok ?

Maybe optionally include the blas stuff?

I wrote it as an option for now.