Bump oldest CUDA 12 cuda-bindings to 12.9.3

[csadorf]

Updates the oldest CUDA 12 cuda-bindings pin to 12.9.3 to match the minimum version required by cudf 26.12.

Closes #8584

[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: 3347c117-e494-4109-8a83-b2153fe28c05

📥 Commits

Reviewing files that changed from the base of the PR and between 7c8c883 and 4017f79.

📒 Files selected for processing (7)

• conda/environments/all_cuda-129_arch-aarch64.yaml
• conda/environments/all_cuda-129_arch-x86_64.yaml
• conda/environments/all_cuda-133_arch-aarch64.yaml
• conda/environments/all_cuda-133_arch-x86_64.yaml
• conda/recipes/cuml/recipe.yaml
• dependencies.yaml
• python/cuml/pyproject.toml

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

📝 Summary

Summary by CodeRabbit

• Chores
  • Updated CUDA 12 dependency requirements to use cuda-bindings 12.9.3.
  • Updated CUDA 13 dependency requirements to use cuda-bindings 13.0.2.
  • Removed the rapidsai Conda channel from dependency configuration.

Walkthrough

The change removes the rapidsai Conda channel and aligns CUDA 12 and CUDA 13 cuda-bindings minimum versions across dependency pins, Conda environments, the Conda recipe, and Python wheel metadata.

Changes

CUDA bindings dependency alignment

Layer / File(s)	Summary
Update test dependency constraints dependencies.yaml	Oldest-dependency and CUDA-specific constraints now require cuda-bindings 12.9.3 for CUDA 12 and 13.0.2 for CUDA 13.
Update Conda dependency constraints conda/environments/*, conda/recipes/cuml/recipe.yaml	CUDA 12 and CUDA 13 Conda environments and recipe host and runtime requirements use the updated minimum versions.
Update Python package constraints python/cuml/pyproject.toml	Runtime and CUDA wheel build requirements now require cuda-bindings>=13.0.2,<14.0.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Severity of issue fixed: Medium

Suggested reviewers: bdice

Merge Risk: ⚪ Minimal · up to 4017f

The update raises cuda-bindings minimum versions consistently so cuML wheel dependency resolution can select compatible CUDA bindings. No remaining merge-readiness risk is identified.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Out of Scope Changes check	⚠️ Warning	The pull request also removes the rapidsai Conda channel and updates CUDA 13 cuda-bindings pins to 13.0.2. These changes are not required by issue #8584, which concerns the CUDA 12 oldest-dependencies…	Remove the rapidsai channel change and CUDA 13 pin updates from this pull request, or provide linked issue requirements that explicitly justify them. Submit unrelated changes separately if they are needed for other issues.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the primary change: updating the oldest CUDA 12 cuda-bindings version to 12.9.3.
Description check	✅ Passed	The description is directly related to the dependency pin update and explains the cuDF compatibility requirement and linked issue.
Linked Issues check	✅ Passed	The pull request updates the CUDA 12 cuda-bindings constraints from 12.9.2 to 12.9.3 across the relevant environments, recipe, and Python package metadata. This addresses the dependency-resolution fai…
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…

Full details: Out of Scope Changes check

Explanation

The pull request also removes the rapidsai Conda channel and updates CUDA 13 cuda-bindings pins to 13.0.2. These changes are not required by issue #8584, which concerns the CUDA 12 oldest-dependencies configuration.

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[csadorf]

All CI failures are unrelated to this change:

• The latest-deps conda jobs fail because the RAPIDS xgboost 3.3.0 packages require an older librmm, which cannot resolve alongside libnvforest 26.12. Tracked in [CI] RAPIDS XGBoost packages block 26.12 conda environments #8583.
• The pip devcontainer jobs fail because the corresponding 26.12 base images are unavailable. Tracked in [CI] 26.12 pip devcontainer builds fail to pull base images #8576.

All relevant oldest-deps coverage passes, including the cuML and cuML Dask wheel tests. This confirms that updating cuda-bindings to 12.9.3 fixes #8584. I recommend admin-merging once the required approval is in.

[jameslamb]

Approving so you have what you need once you've addressed my comments... but I think more is needed here.

[jakirkham]

Thanks Simon and James! 🙏

Have added suggestions for dependencies.yaml below to capture some of James' recommendations

Would still need to update Conda recipes and other files separately (as James noted)

[csadorf]

I’m going to admin-merge this to unblock CI. It’s approved, and all four failing wheel-build jobs are caused by the unrelated libkvikio.so wheel-repair issue tracked in #8603 and fixed by #8604. Merging this will allow #8604 to pass its oldest-deps wheel tests.