DOC reference parallel_config instead of parallel_backend#1457
DOC reference parallel_config instead of parallel_backend#1457tomMoral merged 19 commits intojoblib:masterfrom
Conversation
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## master #1457 +/- ##
==========================================
- Coverage 94.93% 94.83% -0.10%
==========================================
Files 44 45 +1
Lines 7439 7471 +32
==========================================
+ Hits 7062 7085 +23
- Misses 377 386 +9
☔ View full report in Codecov by Sentry. |
Co-authored-by: Jérémie du Boisberranger <34657725+jeremiedbb@users.noreply.github.com>
ogrisel
left a comment
There was a problem hiding this comment.
I did not have time to fully review the PR carefully but what I read looks good.
Co-authored-by: Olivier Grisel <olivier.grisel@ensta.org>
Co-authored-by: Olivier Grisel <olivier.grisel@ensta.org>
|
@tomMoral I do not understand where the doc problems come from when reading the readthedocs output. Do you get more informative feedback when building the doc locally? Readthedocs is also failing on unrelated PRs (e.g. #1463) but it would be great to fix it here to see the updated doc for the |
|
I tried locally and the output folders stay mostly empty but there is no error message and the return code is 0... |
|
So the sphinx crash is probably caused by 2 offending examples:
Deleting both example files makes the sphinx generation complete successfully. But those failures are not related to this PR. I think they also fail on EDIT: I opened #1465 for the first problem. |
|
Actually the above comment is macos only. On the output of the CI it fails because of a warning: |
|
@tomMoral I added the https://joblib--1457.org.readthedocs.build/en/1457/generated/joblib.parallel_config.html |
|
Yeah! thanks @ogrisel :) I don't think the broken build on sklearn is due to this PR as this seems to be an update on scipy\ so merging this one :) |
|
Indeed, i think the scipy solver problem was fixed in scikit-learn main branch and will be released as part of 1.3. |
Follow up for #1451 to use
parallel_configas much as possible in examples and in the tests.In doing so, I realized that it actually has a different behavior than
parallel_backend.This PR does:
default_n_jobsper backend (in particular to have defaultn_jobs=-1for dask).parallel_configusage. Calling the two following examples is equivalent while the first one would giveLokyBackendwith2jobs with the current version onmaster.