Expose fitted attributes for Dask linear models

[nethum529]

Summary

• Expose coef_, intercept_, and n_iter_ from Dask ElasticNet and Lasso.
• Let Dask Ridge proxy coef_ and intercept_ from its fitted model.
• Carry the client output type when a fitted attribute is read from a worker.
• Add coverage for pre-fit errors, output types, and refits.

Testing

• Changed-file pre-commit checks pass.
• The new Dask tests cover ElasticNet, Lasso, and Ridge.

Addresses #7460

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[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: a890f692-4966-4f18-a2e7-f88e086e8193

📥 Commits

Reviewing files that changed from the base of the PR and between af90d6f and b9956ab.

📒 Files selected for processing (6)

• python/cuml/cuml/dask/common/base.py
• python/cuml/cuml/dask/linear_model/elastic_net.py
• python/cuml/cuml/dask/linear_model/lasso.py
• python/cuml/tests/dask/test_dask_base.py
• python/cuml/tests/dask/test_dask_coordinate_descent.py
• python/cuml/tests/dask/test_dask_ridge_regression.py

🚧 Files skipped from review as they are similar to previous changes (1)

• python/cuml/cuml/dask/common/base.py

---

📝 Walkthrough

Summary by CodeRabbit

• New Features

  • Distributed ElasticNet and Lasso now expose fitted properties (coefficients, intercept, and iteration count) through their public attribute access.
  • Improved handling of the active output type when retrieving model attributes in distributed estimators.

• Bug Fixes

  • Pre-fit access to fitted properties now consistently raises an AttributeError indicating the estimator is not fitted.
  • Fitted attributes (including iteration counts) correctly update after refitting for Ridge, ElasticNet, and Lasso.

• Tests

  • Expanded Dask test coverage for fitted-attribute behavior, output-type scenarios, refit correctness, and iteration-count consistency.
  • Added unit tests to ensure attribute-resolution errors are preserved with their original causes/messages.

Walkthrough

Dask linear models now expose fitted attributes through their underlying solvers. Attribute resolution preserves the configured output type, Ridge fitted attributes remain absent before fitting, and tests cover solver consistency, NumPy output, and refit updates.

Changes

Dask fitted attributes

Layer / File(s)	Summary
Output-type-aware attribute lookup python/cuml/cuml/dask/common/base.py, python/cuml/tests/dask/test_dask_base.py	BaseEstimator preserves the active GlobalSettings().output_type while resolving attributes from the internal model and preserves lookup errors.
Solver-backed fitted properties python/cuml/cuml/dask/linear_model/elastic_net.py, python/cuml/cuml/dask/linear_model/lasso.py, python/cuml/cuml/dask/linear_model/ridge.py	ElasticNet and Lasso expose coef_, intercept_, and n_iter_ from their solvers; Ridge no longer initializes fitted attributes before fitting.
Fitted attribute validation python/cuml/tests/dask/test_dask_coordinate_descent.py, python/cuml/tests/dask/test_dask_ridge_regression.py	Tests validate pre-fit absence, solver consistency, output types, iteration counts, and attribute updates after refitting.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related PRs

• rapidsai/cuml#8378: Both changes modify Dask BaseEstimator attribute-resolution behavior.

Suggested labels: improvement, non-breaking

Suggested reviewers: dantegd

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title concisely describes the main change: exposing fitted attributes for Dask linear models.
Description check	✅ Passed	The description clearly matches the changeset by summarizing fitted attributes, output-type handling, and added tests.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[csadorf]

Is the output_type setting required for #7460, or is it a separate improvement? It affects all remote Dask estimators, so I think it needs shared coverage here or should be handled separately. If retained, please use with using_output_type(output_type):.