You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Dec 16, 2022. It is now read-only.
allennlp test-install just runs pytest in a roundabout way, and adds extra maintenance work since we have to make sure this function stays up-to-date with our pytest configuration in Makefile and pytest.ini.
It's also redundant (in our CI at least) to be running the unit tests again, and I don't how useful this command is to end users. In theory, if they (the users) can successfully install AllenNLP on an officially supported OS and Python version, they shouldn't need to run a bunch of unit tests to verify that their installation works. They should trust that we've already done that (and we have).
At most I would suggest they just run allennlp --help to make sure the binary has been installed and torch and what-not can be successfully loaded.
The text was updated successfully, but these errors were encountered:
@epwalsh can you find and link to the PR where we added this? We didn't add it too long ago and it would be good to know why we're circling back on this issue.
Instead of completely removing test-install we could just make it print out some useful info (like version, pytorch version, GPUs available, etc) so that if there are people out there that use allennlp test-install programmatically - like in part of their own CI pipeline - it won't break anything for them.
allennlp test-install
just runspytest
in a roundabout way, and adds extra maintenance work since we have to make sure this function stays up-to-date with our pytest configuration inMakefile
andpytest.ini
.It's also redundant (in our CI at least) to be running the unit tests again, and I don't how useful this command is to end users. In theory, if they (the users) can successfully install AllenNLP on an officially supported OS and Python version, they shouldn't need to run a bunch of unit tests to verify that their installation works. They should trust that we've already done that (and we have).
At most I would suggest they just run
allennlp --help
to make sure the binary has been installed and torch and what-not can be successfully loaded.The text was updated successfully, but these errors were encountered: