Closed brahmaneya closed 3 years ago
(which correctly has a Cython dependency)
For information, the issue not a Cython dependency, but that for python 3.8 users are trying to build scikit-learn 0.22 from sources since binary wheels do not exist for that Python version. And there yes, it looks like Cython should have been declared a build dependency but it wasn't.
BTW, if you want to auto-close the associate issue the PR description needs to say "Fixes #1620" (there is a formatting issue currently).
(which correctly has a Cython dependency)
For information, the issue not a Cython dependency, but that for python 3.8 users are trying to build scikit-learn 0.22 from sources since binary wheels do not exist for that Python version. And there yes, it looks like Cython should have been declared a build dependency but it wasn't.
Building Cython doesn't actually fix the problem, however.
Building Cython doesn't actually fix the problem, however.
You shouldn't need Cython to install a recent version of scikit-learn . What issue are you experiencing? If I install this branch of snorkel on Python 3.8
conda create -n test-env python=3.8 scipy
conda activate test-env
pip install https://github.com/snorkel-team/snorkel/archive/sklearn_bump.zip
it works fine for me:
Merging #1628 (af6f152) into master (ed77718) will not change coverage. The diff coverage is
n/a
.
@@ Coverage Diff @@
## master #1628 +/- ##
=======================================
Coverage 97.21% 97.21%
=======================================
Files 68 68
Lines 2151 2151
Branches 345 345
=======================================
Hits 2091 2091
Misses 31 31
Partials 29 29
As a note: we turned on weekly cron builds to detect issues from changing dependency on an ongoing basis
@henryre thanks!
Description of proposed changes
Build config updates:
Package version updates:
Fixes #1627
Test plan
Checklist
Need help on these? Just ask!
tox -e complex
and/ortox -e spark
if appropriate.