Use CUDA 13.2 in Conda, Devcontainers, Spark, GHA, etc. by jakirkham · Pull Request #2365 · rapidsai/rmm

jakirkham · 2026-04-21T15:49:51Z

Addressing issue:

Update RAPIDS to CUDA 13.2 build-planning#265

Description

Use CUDA 13.2 throughout different build and test environments.

copy-pr-bot · 2026-04-21T15:49:55Z

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.

coderabbitai · 2026-04-21T15:52:51Z

📝 Walkthrough

Summary by CodeRabbit

Release Notes

Chores
- Updated CUDA support to version 13.2 across development environments, CI/CD pipelines, and dependency configurations.
Documentation
- Updated installation instructions to reflect CUDA 13.2 as the standard version.

Walkthrough

This PR updates the project's CUDA version from 13.1 to 13.2 across devcontainer configurations, conda environment specifications, GitHub Actions workflows, dependency matrices, and documentation files.

Changes

Cohort / File(s)	Summary
Devcontainer Configuration `.devcontainer/cuda13.2-conda/devcontainer.json`, `.devcontainer/cuda13.2-pip/devcontainer.json`	Updated CUDA version from 13.1 to 13.2 in build arguments, base image tags, container names, and environment cache/venv directory paths.
Conda Environment Files `conda/environments/all_cuda-132_arch-aarch64.yaml`, `conda/environments/all_cuda-132_arch-x86_64.yaml`	Updated CUDA version from 13.1 to 13.2 in environment dependencies and renamed environments from `all_cuda-131_` to `all_cuda-132_`.
Workflow, Dependencies, and Documentation `.github/workflows/pr.yaml`, `dependencies.yaml`, `README.md`	Updated CUDA matrix input from 13.1 to 13.2 in workflow; updated dependency matrix and package constraints in dependencies.yaml; updated conda install command in README.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

build wheels with CUDA 13.0.x, test wheels against mix of CTK versions #2270: Updates CI/dependency configuration for CUDA version matrix and devcontainer workflows.
ensure 'torch' CUDA wheels are installed in CI #2279: Updates CUDA-versioned CI/dependency configuration across similar configuration files.

Suggested reviewers

jameslamb
bdice

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically summarizes the main change: updating CUDA version from 13.1 to 13.2 across multiple environments and configurations.
Description check	✅ Passed	The description is directly related to the changeset, referencing the specific issue being addressed and explaining that CUDA 13.2 is being deployed across different build and test environments.
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.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

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

coderabbitai

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

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

Inline comments:
In `@README.md`:
- Line 39: Docs are inconsistent: the conda install line was changed to use
cuda-version=13.2 but the build-from-source example still references the old
file name all_cuda-131_arch-$(uname -m).yaml; update the build-from-source
example to the matching CUDA 13.2 name (e.g., rename all_cuda-131_arch-$(uname
-m).yaml to all_cuda-132_arch-$(uname -m).yaml) and verify any other occurrences
of "131" or "131" naming are replaced so the README consistently references CUDA
13.2.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

Push a commit to this branch (recommended)
Create a new PR with the fixes

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: c45f7645-7a69-4bb4-a106-1e6d869add99

📥 Commits

Reviewing files that changed from the base of the PR and between 386f76d and 12613c0.

📒 Files selected for processing (7)

.devcontainer/cuda13.2-conda/devcontainer.json
.devcontainer/cuda13.2-pip/devcontainer.json
.github/workflows/pr.yaml
README.md
conda/environments/all_cuda-132_arch-aarch64.yaml
conda/environments/all_cuda-132_arch-x86_64.yaml
dependencies.yaml

coderabbitai · 2026-04-21T15:52:55Z


 ```bash
-conda install -c rapidsai -c conda-forge rmm cuda-version=13.1
+conda install -c rapidsai -c conda-forge rmm cuda-version=13.2


⚠️ Potential issue | 🟡 Minor

Update the source-build env file example to CUDA 13.2 naming.

The command at Line 39 is updated, but the build-from-source example still references all_cuda-131_arch-$(uname -m).yaml (Line 91). That leaves docs inconsistent and can break setup.

Suggested doc fix

- $ conda env create --name rmm_dev --file conda/environments/all_cuda-131_arch-$(uname -m).yaml + $ conda env create --name rmm_dev --file conda/environments/all_cuda-132_arch-$(uname -m).yaml

📝 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

conda install -c rapidsai -c conda-forge rmm cuda-version=13.2

$ conda env create --name rmm_dev --file conda/environments/all_cuda-132_arch-$(uname -m).yaml

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed. In `@README.md` at line 39, Docs are inconsistent: the conda install line was changed to use cuda-version=13.2 but the build-from-source example still references the old file name all_cuda-131_arch-$(uname -m).yaml; update the build-from-source example to the matching CUDA 13.2 name (e.g., rename all_cuda-131_arch-$(uname -m).yaml to all_cuda-132_arch-$(uname -m).yaml) and verify any other occurrences of "131" or "131" naming are replaced so the README consistently references CUDA 13.2.

Good catch! This is only in the README. AFAICT nowhere else

Will follow up in a bit

🧩 Analysis chain

Let me quickly verify there are no other occurrences of the old 131 CUDA naming in the repo.

🏁 Script executed:

#!/bin/bash rg -n "cuda.?13[._-]?1|all_cuda-131|cuda-131|cuda13\.1|CUDA 13\.1" --type-add 'yaml:*.yaml' -t yaml -t markdown -t json

Length of output: 432

@jakirkham, confirmed! The search results back you up:

README.md line 91 — the only actionable stale reference (the one flagged in my review comment).

CHANGELOG.md lines 8, 97, 98 — historical entries documenting when CUDA 13.1 support was added/fixed; those are correct as-is and shouldn't be touched.

So the fix is just the one-liner in README.md:

- $ conda env create --name rmm_dev --file conda/environments/all_cuda-131_arch-$(uname -m).yaml + $ conda env create --name rmm_dev --file conda/environments/all_cuda-132_arch-$(uname -m).yaml

🧠 Learnings used

Learnt from: jameslamb Repo: rapidsai/rmm PR: 2270 File: ci/test_wheel_integrations.sh:47-48 Timestamp: 2026-03-04T20:02:46.087Z Learning: In rapidsai/rmm (and likely other RAPIDS projects), the CUDA version gate in `ci/test_wheel_integrations.sh` for PyTorch tests is intentionally broad (e.g., all CUDA 12.x and 13.0). It relies on `rapids-dependency-file-generator` to fail loudly when no matching matrix entry exists in `dependencies.yaml` for unsupported CUDA versions. Do not suggest tightening the shell gate to enumerate specific supported versions, as that would duplicate the version list unnecessarily.

Learnt from: jameslamb Repo: rapidsai/rmm PR: 2270 File: dependencies.yaml:457-480 Timestamp: 2026-03-04T00:37:06.930Z Learning: In rapidsai/rmm PR `#2270`, the `depends_on_pytorch` dependency in `dependencies.yaml` pins `torch==2.4.0+cu124` for the CUDA 12.2 + `dependencies: "oldest"` matrix. This version does not have wheels for Python 3.13 or 3.14 on the cu124 index (https://download.pytorch.org/whl/cu124). This is intentionally left as-is because the current CI matrix does not exercise that combination and the nightly matrix does not run integration jobs. A missing wheel for that combination will produce a loud install error if the matrix ever includes Python 3.13/3.14 with CUDA 12.2 oldest dependencies.

jakirkham · 2026-04-21T16:03:34Z

/ok to test

jakirkham added 4 commits April 21, 2026 08:49

Update devcontainers to CUDA 13.2

d5a98ad

Use CUDA 13.2 in Conda envs

978c0bc

Use CUDA 13.2 in GHA

068a210

Use CUDA 13.2 in docs

12613c0

jakirkham added non-breaking Non-breaking change improvement Improvement / enhancement to an existing function labels Apr 21, 2026

github-project-automation Bot added this to RMM Project Board Apr 21, 2026

coderabbitai Bot reviewed Apr 21, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use CUDA 13.2 in Conda, Devcontainers, Spark, GHA, etc.#2365

Use CUDA 13.2 in Conda, Devcontainers, Spark, GHA, etc.#2365
jakirkham wants to merge 4 commits intorapidsai:mainfrom
jakirkham:use_cuda_13.2

jakirkham commented Apr 21, 2026

Uh oh!

copy-pr-bot Bot commented Apr 21, 2026

Uh oh!

coderabbitai Bot commented Apr 21, 2026

Summary by CodeRabbit

Release Notes

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Suggested reviewers

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Apr 21, 2026 •

edited

Loading

Uh oh!

jakirkham Apr 21, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot Apr 21, 2026

Uh oh!

jakirkham commented Apr 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

	conda install -c rapidsai -c conda-forge rmm cuda-version=13.2
	$ conda env create --name rmm_dev --file conda/environments/all_cuda-132_arch-$(uname -m).yaml

Conversation

jakirkham commented Apr 21, 2026

Description

Uh oh!

copy-pr-bot Bot commented Apr 21, 2026

Uh oh!

coderabbitai Bot commented Apr 21, 2026

Summary by CodeRabbit

Release Notes

Walkthrough

Changes

Estimated code review effort

Possibly related PRs

Suggested reviewers

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jakirkham Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

jakirkham commented Apr 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

coderabbitai Bot Apr 21, 2026 •

edited

Loading

jakirkham Apr 21, 2026 •

edited

Loading