Forward merge release/26.04 into main

[jcrist]

No description provided.

[bdice]

Make sure to /merge nosquash when merging this.

[coderabbitai]

📝 Walkthrough

Summary by CodeRabbit

• Documentation

  • Updated scikit-learn compatibility range for cuML accelerator to 1.5–1.8
  • Removed documentation about string array unsupported limitation

• Bug Fixes

  • Improved handling of empty data partitions in distributed processing environments

• Chores

  • Added libopenblas dependency constraint for aarch64 Conda environments

Walkthrough

This PR refactors distributed data handling in Dask workflows to cache per-worker partition row counts and gracefully handle empty partitions. It also updates CI automation, adds OpenBLAS version constraints to conda environments, updates scikit-learn compatibility documentation, and adds comprehensive tests for empty worker scenarios.

Changes

Cohort / File(s)	Summary
CI Workflow Automation .github/workflows/pr_issue_status_automation.yml	Removed the override-item output computation from get-release-version and hardcoded the downstream update-release workflow's OVERRIDE_ITEM input to false instead of deriving it conditionally.
Conda Environment Dependencies conda/environments/all_cuda-129_arch-aarch64.yaml, conda/environments/all_cuda-131_arch-aarch64.yaml	Added libopenblas<=0.3.30 dependency constraint to both aarch64 CUDA conda environments.
Dependency Configuration dependencies.yaml	Added conditional dependency matrix block to pin libopenblas<=0.3.30 for aarch64 conda test environments, placed before the default unconstrained packages matrix.
Documentation Updates docs/source/cuml-accel/limitations.rst	Removed documentation about unsupported string array targets and updated scikit-learn compatibility range from 1.4–1.7 to 1.5–1.8.
Dask Distributed Data Handler Refactoring python/cuml/cuml/dask/common/input_utils.py	Added _worker_sizes caching attribute and _fetch_worker_sizes() method to detect and prune empty workers. Refactored calculate_parts_to_sizes() to use cached worker sizes and removed unused ranks parameter. Emits UserWarning when workers with zero rows are detected.
Dask Nearest Neighbors python/cuml/cuml/dask/neighbors/nearest_neighbors.py	Removed explanatory comment about communicator clique worker membership requirements.
Test Additions python/cuml/tests/dask/test_dask_base.py	Added helper function _make_dask_data_with_empty_worker and three unit tests validating DistributedDataHandler.create warning emission, worker filtering, and row accounting for empty partitions. Added integration test training LogisticRegression on mixed (real + empty) partitions.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• Small update to cuml-accel limitations #7929: Identical documentation updates to scikit-learn compatibility version range and removal of string array limitation.
• ci: preserve Release project field also on release/* PRs #7921: Identical workflow changes removing override-item output and hardcoding OVERRIDE_ITEM to false.
• Fix Dask estimators crashing when workers receive empty partitions #7807: Identical DistributedDataHandler refactoring including worker size caching, signature changes, and empty worker handling tests.

Suggested labels

conda, Cython / Python, improvement, non-breaking, ci, bug

Suggested reviewers

• viclafargue
• csadorf
• msarahan

🚥 Pre-merge checks | ✅ 1 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 58.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	❓ Inconclusive	No description was provided by the author; the pull request lacks any documentation of changes or intent.	Add a pull request description explaining the purpose of this forward merge and summarizing key changes being merged from the release branch.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies this as a forward merge from release/26.04 into main, which accurately describes the pull request's primary purpose.

_{✏️ 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.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@python/cuml/cuml/dask/common/input_utils.py`:
- Around line 145-151: The warnings.warn call in input_utils.py (inside the code
path that logs empty_workers) lacks a stacklevel, causing Ruff B028 and pointing
the warning at the utility instead of the caller; update the warnings.warn
invocation (the one that constructs the message about empty_workers) to include
an explicit stacklevel parameter (e.g., stacklevel=2 or higher) so the warning
originates from the caller location.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 32cac1f0-f724-4434-8bfe-d1da7a44a502

📥 Commits

Reviewing files that changed from the base of the PR and between 5e954f7 and 54d3634.

📒 Files selected for processing (8)

• .github/workflows/pr_issue_status_automation.yml
• conda/environments/all_cuda-129_arch-aarch64.yaml
• conda/environments/all_cuda-131_arch-aarch64.yaml
• dependencies.yaml
• docs/source/cuml-accel/limitations.rst
• python/cuml/cuml/dask/common/input_utils.py
• python/cuml/cuml/dask/neighbors/nearest_neighbors.py
• python/cuml/tests/dask/test_dask_base.py

💤 Files with no reviewable changes (1)

• python/cuml/cuml/dask/neighbors/nearest_neighbors.py

[jcrist]

/merge nosquash

[rapids-bot]

Commit history integrity check failed: not all commits from original PR #7923 appear to be present individually in this PR's history. This usually happens if commits were squashed during the manual resolution process. Please ensure all original commits are preserved individually. You can fix this and try the /merge nosquash command again.