[no-ci] Bump nightly PyTorch highest version to 2.15.0 (cu132 only) by leofang · Pull Request #3031 · NVIDIA/cuda-python

leofang · 2026-10-06T03:30:01Z

Bump the highest nightly-pytorch row in ci/test-matrix.yml from 2.14.1 to 2.15.0 (linux amd64/arm64 and win-64). Torch 2.15.0 ships only as +cu132 on pytorch.org — not cu126 or cu130 — so the two existing 2.14.1 rows per arch (cu126 + cu130) collapse to a single 2.15.0+cu132 row per arch. The 2.9.1 rows (cu126 + cu130) stay as the older anchor.

The 2.15.0+cu132 rows pin CUDA_VER: '13.2.2' to match torch 2.15.0's exact Linux-only dependency on cuda-toolkit==13.2.2; the previous 13.0.2 pin couldn't satisfy that in the install resolver. LOCAL_CTK: '0' on these rows means no mini-CTK artifact for 13.2.2 is required.

The cuda.core tensor-bridge now supports torch 2.15 unconditionally via the stable torch_tensor_from_pyobject shim (#3033).

Nightly matrix was validated pre-merge via a push trigger temporarily added to .github/workflows/ci-nightly.yml and reverted in the same branch. The two commits cancel out — net diff vs main is ci/test-matrix.yml only.

-- Leo's bot

Remove before merging.

copy-pr-bot · 2026-10-06T03:30:05Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

leofang · 2026-10-06T03:32:33Z

/ok to test 6de0ff9

leofang · 2026-10-06T03:45:47Z

/ok to test c720c0d

github-actions · 2026-10-06T04:08:23Z

Doc Preview CI
Preview removed because the pull request was closed or merged.

This reverts commit c720c0d.

PyTorch 2.15.0 ships only as +cu132 on pytorch.org (not cu126/cu130). Replace the two 2.14.1 rows (cu126 + cu130) per arch with a single 2.15.0+cu132 row. The 2.9.1 rows (cu126 + cu130) stay as the older anchor. The cuda.core tensor-bridge supports torch 2.15 unconditionally after the stable PyObject->Tensor shim adoption in NVIDIA#3033.

leofang · 2026-10-31T05:11:08Z

/ok to test 1434c23

PyTorch 2.15.0+cu132 declares an exact Linux dep on cuda-toolkit==13.2.2. The nightly-pytorch install step pins cuda-toolkit[cudart] by matrix TEST_CUDA_MAJOR_MINOR; 13.0.2 (13.0.*) does not admit 13.2.2, so the resolver fails on Linux. Bump CUDA_VER on the three 2.15.0+cu132 rows to 13.2.2 so the pin becomes 13.2.*, which matches torch's requirement. LOCAL_CTK=0 on these rows, so no mini-CTK artifact for 13.2.2 is needed.

leofang · 2026-10-31T17:08:24Z

/ok to test 3198b3e

leofang · 2026-10-31T17:42:02Z

Nightly pipelines passed: https://github.com/NVIDIA/cuda-python/actions/runs/41223063007

This reverts commit 6de0ff9.

coderabbitai · 2026-10-31T17:54:49Z

Summary

Summary by CodeRabbit

Tests
- Nightly test coverage now includes PyTorch 2.15.0 with CUDA 13.2.2 on Linux and Windows for amd64, and Linux for arm64.
- Nightly coverage for PyTorch 2.12.1 with CUDA 12.6.3 and 13.0.2 has been removed from amd64 Linux and Windows. Existing PyTorch 2.9.1 coverage remains.

The nightly PyTorch test matrix adds CUDA 13.2.2 / PyTorch 2.15.0 jobs for Linux and Windows. It removes selected amd64 PyTorch 2.12.1 jobs. Existing PyTorch 2.9.1 jobs remain.

Changes: Nightly test matrix

Layer / File(s)	Summary
Update nightly PyTorch matrix entries `ci/test-matrix.yml`	The Linux matrix adds amd64 and arm64 CUDA 13.2.2 / PyTorch 2.15.0 jobs. The Windows matrix adds an amd64 job with the same versions. Both matrices remove selected amd64 PyTorch 2.12.1 jobs for CUDA 12.6.3 and CUDA 13.0.2.

Suggested reviewers: mdboom

Change: Other

Merge Risk: Minimal · up to 61a77

The jobs have compatible artifacts; only matrix row ordering needs cleanup.

- [ ] Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

coderabbitai

Follow-up Comments

suggestion: Reorder the nightly-pytorch rows by ascending CUDAVER.

103-105: Trivial · Maintainability & Code Quality · Quick win

suggestion: Reorder the nightly-pytorch rows by ascending CUDA_VER.

The current 13.2.2, 12.6.3, 13.0.2 order violates the matrix ordering instruction. Reordering the existing rows is the complete correction and does not change the generated matrix rows.

Suggested fix

-    - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '13.2.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.15.0', TORCH_CUDA: 'cu132' } }
     - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1',  TORCH_CUDA: 'cu126' } }
     - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1',  TORCH_CUDA: 'cu130' } }
+    - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '13.2.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.15.0', TORCH_CUDA: 'cu132' } }
 
-    - { ARCH: 'arm64', PY_VER: '3.12',  CUDA_VER: '13.2.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.15.0', TORCH_CUDA: 'cu132' } }
     - { ARCH: 'arm64', PY_VER: '3.12',  CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1',  TORCH_CUDA: 'cu126' } }
     - { ARCH: 'arm64', PY_VER: '3.12',  CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1',  TORCH_CUDA: 'cu130' } }
+    - { ARCH: 'arm64', PY_VER: '3.12',  CUDA_VER: '13.2.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',     ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.15.0', TORCH_CUDA: 'cu132' } }
 
-    - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '13.2.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'TCC',  ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.15.0', TORCH_CUDA: 'cu132' } }
     - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'TCC',  ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1',  TORCH_CUDA: 'cu126' } }
     - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'TCC',  ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1',  TORCH_CUDA: 'cu130' } }
+    - { ARCH: 'amd64', PY_VER: '3.12',  CUDA_VER: '13.2.2', LOCAL_CTK: '0', GPU: 'l4',         GPU_COUNT: '1', DRIVER: 'latest',  DRIVER_MODE: 'TCC',  ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.15.0', TORCH_CUDA: 'cu132' } }

Open the source location

Review info

Run configuration

Source mode: Auto
Repository: NVIDIA/cuda-python
Branch: main
File: .coderabbit.yaml
Commit: 01037b1
Layers: Repository YAML
Review profile: CHILL
Plan: Enterprise
Run ID: 98da4175-2dc0-4210-a1b1-b3ca73c42e20

Commits

Reviewing files that changed from the base of the PR and between 01037b1 and 61a77b3.

Files selected for processing: 1

ci/test-matrix.yml

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

leofang · 2026-10-31T18:01:42Z

/ok to test 61a77b3

Removed preview folders for the following PRs: - PR #3031 - PR #3222 - PR #3223 - PR #3235 - PR #3236 - PR #3269 - PR #3277 - PR #3294 - PR #3302 - PR #3304 - PR #3307 - PR #3308 - PR #3310 - PR #3331 - PR #3335

leofang added 2 commits October 5, 2026 20:26

Bump nightly PyTorch highest version to 2.14.1

aaeb538

Temporarily add push trigger to ci-nightly.yml for testing

6de0ff9

Remove before merging.

github-actions Bot added the CI/CD CI/CD infrastructure label Oct 6, 2026

cuda.core: bump torch tensor-bridge upper cap to 2.14

c720c0d

github-actions Bot added the cuda.core Everything related to the cuda.core module label Oct 6, 2026

leofang self-assigned this Oct 6, 2026

leofang added this to the cuda.core 1.3.0 milestone Oct 6, 2026

leofang mentioned this pull request Oct 6, 2026

cuda.core: use torch 2.14+ stable PyObject->Tensor shim, drop upper version cap #3033

Merged

leofang added 3 commits October 5, 2026 21:45

Revert "cuda.core: bump torch tensor-bridge upper cap to 2.14"

c43f495

This reverts commit c720c0d.

Merge remote-tracking branch 'origin/main' into bump-torch-2.14.1

6af54d2

leofang changed the title ~~Bump nightly PyTorch highest version to 2.14.1~~ Bump nightly PyTorch highest version to 2.15.0 (cu132 only) Oct 31, 2026

Revert "Temporarily add push trigger to ci-nightly.yml for testing"

61a77b3

This reverts commit 6de0ff9.

leofang changed the title ~~Bump nightly PyTorch highest version to 2.15.0 (cu132 only)~~ [no-ci] Bump nightly PyTorch highest version to 2.15.0 (cu132 only) Oct 31, 2026

leofang marked this pull request as ready for review October 31, 2026 17:47

coderabbitai Bot reviewed Oct 31, 2026

View reviewed changes

leofang requested a review from jpascucci-nv October 31, 2026 18:06

leofang enabled auto-merge (squash) November 1, 2026 18:21

jpascucci-nv approved these changes Nov 2, 2026

View reviewed changes

leofang merged commit 1cda28d into NVIDIA:main Nov 2, 2026
30 checks passed

leofang deleted the bump-torch-2.14.1 branch November 2, 2026 17:06

Navigation Menu

Repository navigation

[no-ci] Bump nightly PyTorch highest version to 2.15.0 (cu132 only) - #3031

[no-ci] Bump nightly PyTorch highest version to 2.15.0 (cu132 only)#3031
leofang merged 8 commits into
NVIDIA:mainfrom
leofang:bump-torch-2.14.1

leofang commented Oct 6, 2026 •

edited

Loading

Uh oh!

copy-pr-bot Bot commented Oct 6, 2026

Uh oh!

leofang commented Oct 6, 2026

Uh oh!

leofang commented Oct 6, 2026

Uh oh!

github-actions Bot commented Oct 6, 2026 •

edited

Loading

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

coderabbitai Bot commented Oct 31, 2026 •

edited

Loading

Summary by CodeRabbit

Uh oh!

coderabbitai Bot left a comment

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

leofang commented Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

copy-pr-bot Bot commented Oct 6, 2026

Uh oh!

leofang commented Oct 6, 2026

Uh oh!

leofang commented Oct 6, 2026

Uh oh!

github-actions Bot commented Oct 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

coderabbitai Bot commented Oct 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Follow-up Comments

Uh oh!

leofang commented Oct 31, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

leofang commented Oct 6, 2026 •

edited

Loading

github-actions Bot commented Oct 6, 2026 •

edited

Loading

coderabbitai Bot commented Oct 31, 2026 •

edited

Loading