wheels: build CUDA 13 wheels with latest CTK (13.3.0)

[jameslamb]

Contributes to rapidsai/build-planning#268

Part of a RAPIDS-wide initiative to build wheels using v13.3.0 of the CUDA Toolkit, and to generally get RAPIDS back to the pattern of always building against its latest supported CTK.

As of rapidsai/shared-workflows#590, RAPIDS builds CUDA 13 wheels with v13.3 of the CTK. This updates the project's nvidia-nvjitlink floor to match that.

[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.

[jameslamb]

/ok to test

[jameslamb]

Failing "Trigger Breaking Change Notifications" is unrelated, tracked here: NVIDIA/cuvs#2309

[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: db5a6512-f8a3-4c30-8b56-9324a11b7e27

📥 Commits

Reviewing files that changed from the base of the PR and between a40f130 and 4889b1a.

📒 Files selected for processing (1)

• python/libraft/pyproject.toml

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

• python/libraft/pyproject.toml

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes
  • Tightened CUDA 13 support by requiring nvidia-nvjitlink version 13.3 or later (still below 14).
  • Updated CUDA Toolkit requirements for CUDA 13 by narrowing which toolkit components are pinned and moving nvidia-nvjitlink to a separate, explicit requirement.
• Chores
  • Updated the project’s SPDX copyright header text.

Walkthrough

CUDA 13 wheel and Python package dependency constraints now require nvidia-nvjitlink 13.3 or newer. Python dependencies separately pin selected CUDA toolkit components to 13.*, and the SPDX copyright attribution is updated.

Changes

CUDA 13 dependency updates

Layer / File(s)	Summary
CUDA 13 wheel dependency constraints dependencies.yaml	The CUDA 13 wheel matrix raises the nvidia-nvjitlink minimum version to 13.3 and reuses the requirement through a YAML anchor.
Python CUDA package requirements python/libraft/pyproject.toml	The package pins selected CUDA toolkit components to version 13.*, declares nvidia-nvjitlink separately with a minimum version of 13.3, and updates the SPDX copyright attribution.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: KyleFromNVIDIA, bdice

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: building CUDA 13 wheels against CUDA Toolkit 13.3.0.
Description check	✅ Passed	The description directly matches the changes, explaining the CUDA 13.3.0 wheel update and the nvjitlink floor bump.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[jameslamb]

I looked at the diff of this PR and thought "huh, wait... shouldn't there be pyproject.toml changes?"

Pushed a change to match what we do in cuVS, where the CUDA 13 dependencies are listed in pyproject.toml as documentation.

https://github.com/NVIDIA/cuvs/blob/8b97b612bec64177ee4e85d1e03654dcb4cb7db7/dependencies.yaml#L433-L438

[jameslamb]

/merge

[jameslamb]

All wheels CI passed, and this change doesn't affect conda in any way. Going to admin-merge this to keep the migration moving (some things need to be merged in dependency order to get packages rebuilt correctly).