[ENH] enable array API support in LogisticRegression by avolkov-intel · Pull Request #2941 · uxlfoundation/scikit-learn-intelex

avolkov-intel · 2026-02-11T17:38:58Z

Description

Add array API support for GPU LogisticRegression
Refactor LogisticRegression estimator, move all checks to sklearnex
Implement separate sklearnex SPMD estimator for LogisticRegression

Checklist:

Completeness and readability

I have commented my code, particularly in hard-to-understand areas.
I have updated the documentation to reflect the changes or created a separate PR with updates and provided its number in the description, if necessary.
Git commit message contains an appropriate signed-off-by string (see CONTRIBUTING.md for details).
I have resolved any merge conflicts that might occur with the base branch.

Testing

I have run it locally and tested the changes extensively.
All CI jobs are green or I have provided justification why they aren't.
I have extended testing suite if new functionality was introduced in this PR.

Performance

I have measured performance for affected algorithms using scikit-learn_bench and provided at least a summary table with measured data, if performance change is expected.
I have provided justification why performance and/or quality metrics have changed or why changes are not expected.
I have extended the benchmarking suite and provided a corresponding scikit-learn_bench PR if new measurable functionality was introduced in this PR.

codecov · 2026-02-11T18:09:59Z

Codecov Report

❌ Patch coverage is 19.40299% with 54 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
sklearnex/linear_model/logistic_regression.py	10.41%	43 Missing ⚠️
onedal/linear_model/logistic_regression.py	42.10%	11 Missing ⚠️

Flag	Coverage Δ
azure	`?`
github	`82.04% <19.40%> (+0.21%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
onedal/linear_model/logistic_regression.py	`29.76% <42.10%> (+4.22%)`	⬆️
sklearnex/linear_model/logistic_regression.py	`49.43% <10.41%> (-9.31%)`	⬇️

... and 30 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

david-cortes-intel · 2026-02-12T15:21:09Z

onedal/linear_model/logistic_regression.py

+    @supports_queue
+    def fit(self, X, y, queue=None):
+
+        # Is sparsity check here fine? - Same in BasicStatistics


Answer is: yes, provided that the sklearnex validation already happened before.

david-cortes-intel · 2026-02-12T15:21:49Z

sklearnex/linear_model/logistic_regression.py

-                dtype=[np.float64, np.float32],
-            )
+            y_proba = self._onedal_predict_proba(X, queue)
+            xp, is_array_api_complient = get_namespace(X)


Suggested change

xp, is_array_api_complient = get_namespace(X)

xp, is_array_api_compliant = get_namespace(X)

david-cortes-intel · 2026-02-12T15:22:45Z

sklearnex/linear_model/logistic_regression.py

+                min_prob = 1e-15
+                max_prob = 1.0 - 1e-15
+
+            # TODO ARRAY API only branch check how this code can be adapted


What do you mean "adapted"? This already works with array API, and there is a scikit-learn method that can work on CPU with the fitted coefficients from oneDAL when they are on CPU.

david-cortes-intel · 2026-02-12T15:23:09Z

sklearnex/linear_model/logistic_regression.py

-                accept_large_sparse=_sparsity_enabled,
-                dtype=[np.float64, np.float32],
-            )
+            xp, is_array_api_complient = get_namespace(X)


Suggested change

xp, is_array_api_complient = get_namespace(X)

xp, is_array_api_compliant = get_namespace(X)

david-cortes-intel · 2026-02-12T15:24:44Z

sklearnex/linear_model/logistic_regression.py


        def _onedal_predict(self, X, queue=None):
            if queue is None or queue.sycl_device.is_cpu:
+                #TODO modify function to return array api compliant results???


No, if the input is not supported, it should be offloaded to scikit-learn, which will likely error out on its own. Either way, daal4py makes such a check for whether to offload to scikit-learn or not:

scikit-learn-intelex/daal4py/sklearn/linear_model/logistic_path.py

Line 441 in bf43701

if not _dal_ready:

So it's safe to not modify the output here.

david-cortes-intel · 2026-02-12T15:29:31Z

sklearnex/linear_model/logistic_regression.py


        def _onedal_score(self, X, y, sample_weight=None, queue=None):
+            # TODO1 - OK 
+            # is accuracy_score array api compatible? Seems it is functions from skelarn it's ARRAY API compatible 


Yes, you can check for compatibility in their documentation:
https://scikit-learn.org/stable/modules/array_api.html#metrics

david-cortes-intel · 2026-02-12T15:31:00Z

onedal/linear_model/logistic_regression.py

+
+        # TODO do we need to support behavior when model fitted with sklearn 
+        # (e.g. torch tensor or else and then this method is run)
+        # Currently it can't 


It should be possible, because the oneDAL model object would only require coefficients and intercepts.

david-cortes-intel · 2026-02-12T15:32:26Z

onedal/linear_model/logistic_regression.py


        self._onedal_model = result.model
+
+        # For now it's fine to keep n_iteration as numpy variable 


I guess the answer should be yes for scikit-learn compatibility, since they do not fully support array API for logistic regression at the moment. But that might change in the future.

david-cortes-intel · 2026-02-12T15:32:49Z

onedal/linear_model/logistic_regression.py

+        # Is sparsity check here fine? - Same in BasicStatistics
        is_csr = _is_csr(X)

+        # Is it good place? - Same in LinReg


Yes, but remember that this attribute also needs to be present in the sklearnex object, because sklearn has it.

david-cortes-intel · 2026-02-12T15:34:41Z

sklearnex/linear_model/logistic_regression.py

-                dtype=[np.float64, np.float32],
-            )
+
+            xp, is_array_api_complient = get_namespace(X)


Suggested change

xp, is_array_api_complient = get_namespace(X)

xp, is_array_api_compliant = get_namespace(X)

david-cortes-intel · 2026-02-12T15:38:13Z

@avolkov-intel This will also require updates to the documentation for array API here:
https://uxlfoundation.github.io/scikit-learn-intelex/2025.10/array_api.html

Since LogisticRegression will be a special case that won't work like the rest and won't meet some of the specifications mentioned in that doc - for example:

avolkov-intel added 2 commits February 11, 2026 07:06

Initial refactoring of LogReg

f6845a6

Align Logistic Regression SPMD estimator with LinReg SPMD estimator

7723bdd

avolkov-intel added enhancement New feature or request Array API labels Feb 11, 2026

david-cortes-intel reviewed Feb 12, 2026

View reviewed changes

Update spmd example

5cef2b8

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[ENH] enable array API support in LogisticRegression#2941

[ENH] enable array API support in LogisticRegression#2941
avolkov-intel wants to merge 3 commits intouxlfoundation:mainfrom
avolkov-intel:dev/logreg-array-api

avolkov-intel commented Feb 11, 2026

Uh oh!

codecov bot commented Feb 11, 2026 •

edited

Loading

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel Feb 12, 2026

Uh oh!

david-cortes-intel commented Feb 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	xp, is_array_api_complient = get_namespace(X)
	xp, is_array_api_compliant = get_namespace(X)


		self._onedal_model = result.model

		# For now it's fine to keep n_iteration as numpy variable

Conversation

avolkov-intel commented Feb 11, 2026

Description

Uh oh!

codecov bot commented Feb 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

david-cortes-intel commented Feb 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

codecov bot commented Feb 11, 2026 •

edited

Loading