Test cuml.accel upstream test suite with sklearn 1.9

[csadorf]

Updates the cuml.accel upstream scikit-learn test target matrix to cover scikit-learn 1.9.0, and updates the xfail list for test IDs and behavior changes across the supported 1.6, 1.8, and 1.9 targets.

Summary

• Bumps the latest upstream scikit-learn accel test target from 1.8.0 to 1.9.0.
• Moves the intermediate target from 1.7.2 to 1.8.0.
• Keeps the oldest target pinned at 1.6.0.
• Scopes existing xfails to the scikit-learn versions where they still apply, avoiding stale xfails on newer targets.
• Adds new scikit-learn>=1.9 xfail groups for renamed tests and behavior differences observed in the updated upstream test suite.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[csadorf]

Codex-generated triage of the newly added xfail groups:

P0: must fix
None from this batch. I don't see a new xfail group that blocks ordinary cuml.accel usage or indicates a clear public API regression.

P1: should fix

• Coordinate descent sample_weight behavior differs under cuml.accel in scikit-learn 1.9.
• LogisticRegression callback/refit behavior differs under cuml.accel in scikit-learn 1.9.
• Sparse/dense coordinate-descent equality has numerical tolerance differences.
• RandomForest max_samples validation differs.
• RandomForest private class/sample-weight helper state differs, adjacent to [FEA] Add class_weight + sample_weight to RandomForest estimators #8093.
• SVC probability deprecation warning behavior differs.
• TargetEncoder warning behavior differs.
• Sparse int64-index validation differs.
• LogisticRegression warm_start with newton solvers differs.

P2: nice-to-have

• HTML representation failures under ASCII-locale CI.
• Missing-value validation message wording differs.
• RandomForest.estimators_samples_ is not exposed under cuml.accel, tracked by [Tracker] Missing attributes in cuml.accel #6966.
• Non-strict SciPy 1.16 deprecation warning noise.
• Float32-vs-float64 precision mismatch in one LogisticRegressionCV scoring test.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 1281500c-2526-49fd-8f10-ee697052dc86

📥 Commits

Reviewing files that changed from the base of the PR and between 8c1f46c and 2f2213b.

📒 Files selected for processing (2)

• dependencies.yaml
• python/cuml/cuml_accel_tests/upstream/scikit-learn/xfail-list.yaml

---

📝 Walkthrough

Summary by CodeRabbit

• Chores
  • Updated scikit-learn test dependency matrix versions: latest to 1.9.0 and intermediate to 1.8.0
  • Refined test exclusion rules with version-specific conditions to address known behavior differences across scikit-learn 1.7.2–1.9 releases

Walkthrough

Updated scikit-learn test matrix pins from 1.8.0/1.7.2 to 1.9.0/1.8.0 and reorganized xfail test exclusion rules across multiple version ranges to accommodate compatibility differences in coordinate descent, ensemble voting, linear models, and numeric precision handling between scikit-learn versions.

Changes

scikit-learn 1.8 and 1.9 compatibility

Layer / File(s)	Summary
Dependency version pins update dependencies.yaml	Updated test_python_accel_sklearn matrix pins to target scikit-learn 1.9.0 for "latest" and 1.8.0 for "intermediate" dependency variants.
Test xfail rules for sklearn version ranges python/cuml/cuml_accel_tests/upstream/scikit-learn/xfail-list.yaml	Reorganized xfail exclusion rules across sklearn version ranges: added new sklearn>=1.9 xfail blocks for HTML repr, LogisticRegression behavior, coordinate descent tolerance, and RandomForest/SVC warnings; split existing xfails (ElasticNet, LogisticRegression, warm-start solvers, SciPy deprecation warnings) into version-specific subranges (>=1.8,<1.9 vs >=1.9); narrowed ROC AUC legend-label cases to >=1.7.2,<1.9; and updated test parameter variants based on sklearn version.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• rapidsai/cuml#8126: scikit-learn 1.9 pre-release test and xfail adjustments including warning suppression and sklearn-version conditional skips.
• rapidsai/cuml#8208: Updates to the same dependencies.yaml test matrix entries defining scikit-learn version pins for test_python_accel_sklearn.
• rapidsai/cuml#8023: Updates to xfail-list.yaml for RandomForest ensemble compatibility changes driven by upstream scikit-learn behavior modifications.

Suggested labels

ci, cuml-accel, sklearn-api-compat

Suggested reviewers

• viclafargue
• dantegd

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The pull request title accurately summarizes the main change: updating cuml.accel to test against scikit-learn 1.9, which is the core objective reflected in both the dependencies.yaml and xfail-list.yaml changes.
Description check	✅ Passed	The pull request description is comprehensive and directly related to the changeset, explaining the dependency matrix bumps (1.8.0→1.9.0), version scoping of xfails, and new xfail groups for sklearn 1.9 behavior changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[jcrist]

/merge