build wheels with CUDA 13.0.x, test wheels against mix of CTK versions

[jameslamb]

Description

Contributes to rapidsai/build-planning#257

• builds CUDA 13 wheels with the 13.0 CTK
• ensures wheels ship with a runtime dependency of nvidia-nvjitlink>={whatever-minor-version-they-were-built-against}

Contributes to rapidsai/build-planning#256

• updates wheel tests to cover a range of CTK versions (we previously, accidentally, were only testing the latest 12.x and 13.x)

Contributes to rapidsai/build-planning#205

• switches rapidsai/shared-workflows branches, @python-3.14 -> @main (feat: add testing for Python3.14 rapidsai/shared-workflows#508 was merged)

Issue

N/A

Checklist

• I am familiar with the Contributing Guidelines.
• Testing
  • New or existing tests cover these changes
• Documentation
  • The documentation is up to date with these changes

[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

Updates GitHub Actions workflows to use main branch versions instead of python-3.14 tags, adds pip constraint handling to CI test scripts, expands CUDA version support matrices in dependencies.yaml, and refactors pyproject.toml to use cuda-toolkit with specific NVIDIA components instead of individual packages.

Changes

Cohort / File(s)	Summary
GitHub Actions Workflow Updates .github/workflows/build.yaml, .github/workflows/pr.yaml	Updated action references from python-3.14 tag to main branch across multiple jobs, removing version pin and adopting latest main-stable definitions. Control flow and error handling unchanged.
CI Test Script Enhancements ci/test_self_hosted_service.sh, ci/test_wheel_cuopt.sh, ci/test_wheel_cuopt_server.sh	Added pip constraint generation via rapids-generate-pip-constraints, updated install commands with --prefer-binary and --constraint flags, extended wheel selection with test extras and additional packages (cuopt_sh_client, libcuopt wheels).
CUDA Dependencies Configuration dependencies.yaml	Replaced CUDA 12.0/12.4 with expanded versions (12.2, 12.5, 12.8, 12.9, 13.0, 13.1); added new requirements-output block with cuda_version-specific matrices and error handling for unsupported CUDA values; refactored CUDA 13 dependencies using anchors (&ctk_cu13, &cudss_cu13, &nvjitlink_cu13) and introduced nvjitlink/cudss components.
Project Dependencies python/libcuopt/pyproject.toml	Replaced individual NVIDIA runtime packages (nvidia-cublas, nvidia-cudart, nvidia-cudss, nvidia-curand, nvidia-cusolver, nvidia-cusparse, nvidia-nvjitlink, nvidia-nvtx) with cuda-toolkit [cublas,cudart,curand,cusolver,cusparse,nvtx]==13.\*, nvidia-cudss-cu13, and nvidia-nvjitlink >=13.0,<14.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title directly and clearly summarizes the main changes: building CUDA 13.0.x wheels and testing against multiple CTK versions, which aligns with the primary objectives of the pull request.
Description check	✅ Passed	The description is well-related to the changeset, providing context for the three main objectives: CUDA 13 wheel builds, expanded CTK version testing, and workflow branch updates.
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.}

[rgsl888prabhu]

@jameslamb I have moved this 26.04 release branch, please change it if it is not meant for that.

[jameslamb]

@jameslamb I have moved this 26.04 release branch, please change it if it is not meant for that.

Thanks! Yep I intended this for 26.04, there just wasn't a release/26.04 branch yet when I opened it.

[jameslamb]

@rgsl888prabhu is there an issue or PR tracking some of these test segfaults I'm seeing?

e.g.

[ RUN      ] level0_retail/retail_float_test_t.CVRPTW_Retail/4
ROUTING_TEST: /tmp/conda-bld-output/bld/rattler-build_libmps-parser/work/cpp/src/routing/local_search/vrp/vrp_execute.cu:466: bool cuopt::routing::detail::execute_vrp_moves(solution_t<i_t, f_t, REQUEST>&, move_candidates_t<i_t, f_t>&, i_t) [with i_t = int; f_t = float; cuopt::routing::request_t REQUEST = cuopt::routing::request_t::VRP]: Assertion `cost_before - cost_after > EPSILON&& "Cost should improve!"' failed.
./ci/run_ctests.sh: line 24:  5951 Aborted                 (core dumped) "${gt}" "$@"

(conda-cpp-tests)

The C++ tests should be totally unrelated to changed in this PR, so I suspect there are some other issues.

I see those issues in other PRs here too.

[jameslamb]

Those test failures didn't re-occur on a re-run, and I strongly suspect that they were not related to these changes given that several were for C++ and Python conda packages.

I'm confident that we can merge this.

[jameslamb]

/merge