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

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR pins seven wheel-build CI jobs in build.yaml and pr.yaml to the shared-workflows wheels-build.yaml@newest-ctk ref instead of @main, and tightens the nvidia-nvjitlink version constraint to >=13.3,<14 in dependencies.yaml and python/libcudf/pyproject.toml (also updating a copyright line).

Changes

CI Workflow Reference Update

Layer / File(s)	Summary
Wheel-build job ref pin .github/workflows/build.yaml, .github/workflows/pr.yaml	Seven wheel-build jobs (libcudf, libcudf-streaming, cudf-streaming, pylibcudf, cudf, dask-cudf, cudf-polars) updated to use wheels-build.yaml@newest-ctk instead of @main in both workflow files.

Dependency Version Update

Layer / File(s)	Summary
nvjitlink constraint tightening dependencies.yaml, python/libcudf/pyproject.toml	The nvidia-nvjitlink version constraint for CUDA 13 wheels is raised from >=13.0,<14 to >=13.3,<14; copyright line updated to include "AFFILIATES".

Estimated code review effort: 1 (Trivial) | ~5 minutes

Possibly related PRs

• rapidsai/cudf#22825: Modifies the same wheel-build job uses: references in build.yaml/pr.yaml to pin wheels-build.yaml to a different ref (@cuda-13.3.0).
• rapidsai/cudf#22960: Modifies the same uses: references for shared wheel-build workflows in build.yaml.

Suggested labels: libcudf

Suggested reviewers: bdice, KyleFromNVIDIA

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
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.
Title check	✅ Passed	The title clearly matches the main change: updating CUDA 13 wheel builds to the latest CTK 13.3.0.
Description check	✅ Passed	The description accurately describes the CUDA 13 wheel and nvjitlink floor update reflected in the changeset.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[jameslamb]

/merge

[jameslamb]

polars tests are failing in some ways that implicate pyarrow

FAILED py-polars/tests/unit/io/test_ipc.py::test_ipc_roundtrip_pyarrow_parametric[lz4] - FutureWarning: pyarrow.feather.read_table is deprecated as of 24.0.0. Use pyarrow.ipc.open_file() / RecordBatchFileReader instead. Feather V2 is the Arrow IPC file format.
Falsifying example: test_ipc_roundtrip_pyarrow_parametric(
    compression='lz4',
    df=shape: (0, 1)
    ┌──────┐
    │ col0 │
    │ ---  │
    │ i8   │
    ╞══════╡
    └──────┘,  # or any other generated value
)

(build link)

Think we'll need #23218 for CI to pass here.

[jameslamb]

Think we'll need #23218 for CI to pass here.

That's now merged. Updated to latest main here to restart CI. If it passes, the PR should auto-merge from my earlier /merge comment.

Hopefully end-to-end time should be better, now that we have some CTK 13.3 wheel builds already cached via sccache.