[ENH] sklearn 1.6.dev0 adjustments. #335
Conversation
goraj
changed the title
for more information, see https://pre-commit.ci
for more information, see https://pre-commit.ci
goraj
changed the title
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #335 +/- ##
==========================================
- Coverage 80.50% 80.33% -0.18%
==========================================
Files 24 24
Lines 2334 2339 +5
Branches 339 339
==========================================
Hits 1879 1879
- Misses 318 322 +4
- Partials 137 138 +1 ☔ View full report in Codecov by Sentry. |
|
Nothing? |
There was a problem hiding this comment.
Sorry for the delay @goraj and thanks for the PR!
I left a review. So I was thinking, should we just revamp how we're using the parametrize_with_checks function. Instead of labeling sample_weight_equivalence as an ignored test per area, perhaps let's introduce a test_common.py function under treeple/tests/, and then we can do the "sklearn compatible" check there. Then, we can consolidate what tests we want to ignore in the same file.
Kind of like here:
and how expected_failed_checks kwarg of parametrize_with_checks is used to ignore tests inside https://github.com/scikit-learn/scikit-learn/blob/66270e46b77d6202559bae4929ec83ab320beb1e/sklearn/tests/test_common.py#L118
WDYT?
| with parallel_config("multiprocessing"): | ||
| out = Parallel(n_jobs=n_jobs)( | ||
| delayed(_parallel_build_null_forests)( | ||
| y_pred_ind_arr, | ||
| n_estimators, | ||
| all_y_pred, | ||
| y_test, | ||
| seed, | ||
| metric, | ||
| **metric_kwargs, | ||
| ) | ||
| for i, seed in zip(range(n_repeats), ss.spawn(n_repeats)) |
There was a problem hiding this comment.
Why was this change made?
There was a problem hiding this comment.
If I remember correctly the default loky would segfault during unit testing the *Oblique trees.
| with parallel_config("multiprocessing"): | ||
| out = Parallel(n_jobs=n_jobs)( |
Thank you. |
Hi all,
New to the project but I've been loosely following @adam2392 and the project for a while now.
I setup a dev environment according to DEVELOPMENT.md and ran into a few issues due to sklearn 1.6.dev0 being installed. Namely the introduction of
check_sample_weight_equivalencein scikit-learn/scikit-learn@364cafe leads to expected but not skipped test-case failures.What does this implement/fix? Explain your changes.
Changes will skip check_sample_weight_equivalence testing for forest implementations. It also addresses some un-pickling issues encountered during testing due to joblib/loky in the structure of treeple.stats.utils.
Changes should be backward compatible.