Issue Station
Sign In
Tools hereToolbox

pytorch - ✅(Solved) Fix Enable CUDA 13.2 binaries [3 pull requests, 1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
pytorch/pytorch#177067•Fetched 2026-04-08 00:22:21
View on GitHub
Comments
1
Participants
2
Timeline
101
Reactions
6
Author
tinglvvtinglvv
Participants
johnnynunezjohnnynuneztinglvvtinglvv
Assignees
tinglvvtinglvv
Timeline (top)
subscribed ×36mentioned ×34referenced ×17cross-referenced ×7

Fix Action

Fixed

PR fix notes

PR #177083: Add CUDA installation for CUDA 13.2 and almalinux image for magma

Description (problem / solution / changelog)

https://github.com/pytorch/pytorch/issues/177067

Step 1 add almalinux image for magma and installation path for CUDA 13.2.

For future, we should consider removing 12.6 builds to decrease the load on CI/CD, if they are not used anymore.

cc @atalman @malfet @ptrblck @nWEIdia

Changed files

  • .ci/docker/almalinux/Dockerfile (modified, +5/-1)
  • .ci/docker/common/install_cuda.sh (modified, +21/-1)
  • .github/workflows/build-almalinux-images.yml (modified, +1/-1)

PR #7822: Add Windows AMI for CUDA 13.2

Description (problem / solution / changelog)

https://github.com/pytorch/pytorch/issues/177067

Also updates cudnn to 9.19.0.56 to be in sync with linux builds

cc @atalman

Changed files

  • aws/ami/windows/scripts/Installers/Install-CUDA-Tools.ps1 (modified, +6/-1)
  • aws/ami/windows/windows.pkr.hcl (modified, +7/-0)

PR #177197: Add magma build for cuda 13.2

Description (problem / solution / changelog)

Add magma build for CUDA 13.2 after almalinux docker is available

https://github.com/pytorch/pytorch/issues/177067

cc @atalman

Changed files

  • .ci/magma/Makefile (modified, +7/-0)
  • .github/workflows/build-magma-linux.yml (modified, +1/-1)
RAW_BUFFERClick to expand / collapse

🚀 The feature, motivation and pitch

CUDA 13.2 is released on 3/9, creating issue tracker for CUDA 13.2 binaries enablement.

New features in 13.2:

  • NVIDIA CUDA Tile is now supported on devices of compute capability 8.X architectures (NVIDIA Ampere and NVIDIA Ada), as well as 10.X and 12.X architectures (NVIDIA Blackwell)
  • Unified CUDA Toolkit for Arm targets
  • Support for recursive functions and closures in cuTile Python
  • Support for emulated FP64 in math libraries
  • Numba-CUDA debugging to developer tools
  • CUDA Graphs to cuda.core

See https://developer.nvidia.com/blog/cuda-13-2-introduces-enhanced-cuda-tile-support-and-new-python-features/ for more comprehensive notes.

Update Plan

FeatureTarget Date
Linux and Linux aarch64 CD NightlyMar 13, 2026
Enablement of CUDA 13.2 in PyTorch CIMar 20, 2026
Windows CD NightlyMar 20, 2026
Enable CUDA 13.2 in Ecosystem LibrariesMar 24, 2026

PR

Linux builds CD nightly (x86, sbsa, libtorch)

  1. Add Almalinux docker --> Magma build
  1. Builds addition
  2. Windows builds

Additional context

No response

cc @seemethere @malfet @atalman @nWEIdia @ptrblck @msaroufim @eqy @jerryzh168

extent analysis

Fix Plan

Update PyTorch to Support CUDA 13.2

Step 1: Update Docker Images

Update the Almalinux docker image in the Magma build to use CUDA 13.2.

# magma/build.sh
# Update CUDA version
export CUDA_VERSION=13.2

Step 2: Update Linux Builds

Update the Linux builds to use the new CUDA version.

# magma/build.sh
# Update CUDA version
export CUDA_VERSION=13.2
# Update build command
./build.sh --cuda-version=$CUDA_VERSION

Step 3: Update Windows Builds

Update the Windows builds to use the new CUDA version.

# test-infra/build.sh
# Update CUDA version
export CUDA_VERSION=13.2
# Update build command
./build.sh --cuda-version=$CUDA_VERSION

Step 4: Update CI Pipelines

Update the CI pipelines to use the new CUDA version.

# .github/workflows/ci.yml
# Update CUDA version
- name: Update CUDA version
  run: |
    export CUDA_VERSION=13.2
    # Update build command
    ./build.sh --cuda-version=$CUDA_VERSION

Step 5: Verify the Fix

Verify that the PyTorch build is successful with CUDA 13.2.

# Verify build success
if [ $? -eq 0 ]; then
  echo "Build successful with CUDA 13.2"
else
  echo "Build failed with CUDA 13.2"
  exit 1
fi

Extra Tips

  • Make sure to update all relevant dependencies and libraries to use the new CUDA version.
  • Verify that the update does not break any existing functionality.
  • Consider adding tests to ensure that the update is working correctly.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#ssr#installation#tensor shape#autograd error#embedding generation#cache error#pipeline error#runtime error#dependency conflict

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING