Forward merge from release/26.04

[jcrist]

Fixes merge conflicts in #7905.

[coderabbitai]

📝 Walkthrough

Summary by CodeRabbit

• Chores

  • Updated minimum dependency versions: SciPy (1.14.0), scikit-learn (1.5), and treelite (4.6.1+).
  • Set upper bound constraint for numba-cuda to <0.29.0.
  • Removed deprecated test marker.

• Documentation

  • Updated supported versions documentation to reflect new minimum library requirements.

Walkthrough

This PR updates dependency version constraints across conda environments, recipes, and Python project configurations. Key changes include adding an upper bound to numba-cuda (<0.29.0), bumping scipy minimum from >=1.13.0 to >=1.14.0, and updating scikit-learn and treelite versions in documentation. Additionally, the pytest marker no_bad_cuml_array_check is removed.

Changes

Cohort / File(s)	Summary
Conda environment files conda/environments/all_cuda-129_arch-aarch64.yaml, conda/environments/all_cuda-129_arch-x86_64.yaml, conda/environments/all_cuda-131_arch-aarch64.yaml, conda/environments/all_cuda-131_arch-x86_64.yaml	Updated numba-cuda constraint to add upper bound <0.29.0 and bumped scipy minimum to >=1.14.0 across all environment specifications.
Conda recipe conda/recipes/cuml/recipe.yaml	Added upper bound <0.29.0 to numba-cuda in run requirements and increased scipy minimum from >=1.11.0 to >=1.14.0.
Root dependency configuration dependencies.yaml	Narrowed numba-cuda ranges across all CUDA variants (cu12, cu13) by adding <0.29.0 upper bound and raised scipy minimum to >=1.14.0.
Documentation docs/source/supported_versions.rst	Updated runtime dependency versions: scikit-learn >=1.4→>=1.5, scipy >=1.8.0→>=1.14.0, and treelite ==4.4.1→>=4.6.1,<5.0.0.
Python project configuration python/cuml/pyproject.toml	Removed pytest marker no_bad_cuml_array_check and tightened numba-cuda with <0.29.0 upper bound; updated scipy to >=1.14.0 in project dependencies.
Test configuration python/cuml/tests/test_incremental_pca.py	Removed @pytest.mark.no_bad_cuml_array_check decorators from test_fit and test_partial_fit methods.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

• Cap numba-cuda upper bound at <0.29.0 #7900: Directly adds the same upper bound constraint (<0.29.0) to numba-cuda across environment and project files.
• Bump scipy minimum required version to 1.14.0. #7857: Updates the same SciPy minimum version pin to >=1.14.0 across configuration and documentation files.
• refactor: build wheels and conda packages using Python limited API #7758: Modifies conda/recipes/cuml/recipe.yaml build requirements alongside refactoring Python ABI expressions.

Suggested reviewers

• jameslamb
• viclafargue

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'Forward merge from release/26.04' accurately describes the primary purpose of the pull request as a forward merge from a release branch.
Description check	✅ Passed	The description 'Fixes merge conflicts in #7905' is directly related to the changeset, which updates dependencies and removes test markers as part of resolving merge conflicts.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

📝 Coding Plan

• Generate coding plan for human review comments

---

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

Tip

You can make CodeRabbit's review stricter and more nitpicky using the `assertive` profile, if that's what you prefer.

Change the reviews.profile setting to assertive to make CodeRabbit's nitpick more issues in your PRs.

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

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

Inline comments:
In `@conda/recipes/cuml/recipe.yaml`:
- Around line 102-105: Update the scikit-learn minimum version in the conda
recipe by changing the dependency string "scikit-learn >=1.4" to match the
documented constraint "scikit-learn >=1.5"; locate the dependency list in the
recipe.yaml where the dependencies include "numba-cuda", "numpy",
"scikit-learn", and "scipy" and replace the scikit-learn version specifier so it
aligns with pyproject.toml, dependencies.yaml, and supported_versions.rst.

In `@docs/source/supported_versions.rst`:
- Line 18: Update the documented treelite minimum version string that currently
reads "treelite: >=4.6.1,<5.0.0" to match the actual package constraints
"treelite: >=4.7.0,<5.0.0"; verify against the authoritative constraints in
pyproject.toml and dependencies.yaml and ensure the entry in
supported_versions.rst (the treelite line) is edited accordingly so docs and
package constraints are consistent.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 5ea02902-6d45-43f5-84fc-f0c8c7cff7b8

📥 Commits

Reviewing files that changed from the base of the PR and between 95b7be3 and bf7692a.

📒 Files selected for processing (9)

• conda/environments/all_cuda-129_arch-aarch64.yaml
• conda/environments/all_cuda-129_arch-x86_64.yaml
• conda/environments/all_cuda-131_arch-aarch64.yaml
• conda/environments/all_cuda-131_arch-x86_64.yaml
• conda/recipes/cuml/recipe.yaml
• dependencies.yaml
• docs/source/supported_versions.rst
• python/cuml/pyproject.toml
• python/cuml/tests/test_incremental_pca.py

💤 Files with no reviewable changes (1)

• python/cuml/tests/test_incremental_pca.py

[coderabbitai]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Compare sklearn minimum constraints across the files provided in this PR context
rg -n 'scikit-learn' \
  conda/recipes/cuml/recipe.yaml \
  dependencies.yaml \
  python/cuml/pyproject.toml \
  docs/source/supported_versions.rst

Repository: rapidsai/cuml

Length of output: 533

---

Update scikit-learn minimum to match documented constraint.

Line 104 specifies scikit-learn >=1.4 while the documentation and all other dependency definitions (pyproject.toml, dependencies.yaml, supported_versions.rst) require scikit-learn >=1.5. This inconsistency allows conda environments below the tested minimum.

Proposed fix

-    - scikit-learn >=1.4
+    - scikit-learn >=1.5

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	- numba-cuda >=0.22.2,<0.29.0
	- numpy >=1.23,<3.0
	- scikit-learn >=1.4
	- scipy >=1.11.0
	- scipy >=1.14.0
	- numba-cuda >=0.22.2,<0.29.0
	- numpy >=1.23,<3.0
	- scikit-learn >=1.5
	- scipy >=1.14.0

🤖 Prompt for AI Agents

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

In `@conda/recipes/cuml/recipe.yaml` around lines 102 - 105, Update the
scikit-learn minimum version in the conda recipe by changing the dependency
string "scikit-learn >=1.4" to match the documented constraint "scikit-learn
>=1.5"; locate the dependency list in the recipe.yaml where the dependencies
include "numba-cuda", "numpy", "scikit-learn", and "scipy" and replace the
scikit-learn version specifier so it aligns with pyproject.toml,
dependencies.yaml, and supported_versions.rst.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# Verify treelite constraints in docs vs packaging metadata/source-of-truth
rg -n 'treelite' docs/source/supported_versions.rst \
  dependencies.yaml \
  python/cuml/pyproject.toml \
  conda/recipes/cuml/recipe.yaml

Repository: rapidsai/cuml

Length of output: 504

---

Align treelite minimum in docs with the actual package floor.

Line 18 documents treelite>=4.6.1,<5.0.0, but the actual package constraints are >=4.7.0,<5.0.0 (per pyproject.toml and dependencies.yaml). This mismatch can direct users to unsupported versions.

Proposed fix

-* **treelite**: >=4.6.1,<5.0.0
+* **treelite**: >=4.7.0,<5.0.0

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	* **treelite**: >=4.6.1,<5.0.0
	* **treelite**: >=4.7.0,<5.0.0

🤖 Prompt for AI Agents

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

In `@docs/source/supported_versions.rst` at line 18, Update the documented
treelite minimum version string that currently reads "treelite: >=4.6.1,<5.0.0"
to match the actual package constraints "treelite: >=4.7.0,<5.0.0"; verify
against the authoritative constraints in pyproject.toml and dependencies.yaml
and ensure the entry in supported_versions.rst (the treelite line) is edited
accordingly so docs and package constraints are consistent.

[gforsyth]

Admin merging this to get main up-to-date while upstream packages are broken / not building