Skip to content

chore: upgrade cuda 12.8 for pytorch 2.8#218

Draft
KeijiBranshi wants to merge 3 commits into
mainfrom
fix/upgrade-cuda-12.8-for-pytorch-2.8
Draft

chore: upgrade cuda 12.8 for pytorch 2.8#218
KeijiBranshi wants to merge 3 commits into
mainfrom
fix/upgrade-cuda-12.8-for-pytorch-2.8

Conversation

@KeijiBranshi

Copy link
Copy Markdown
Member

No description provided.

KeijiBranshi and others added 3 commits April 15, 2026 23:12
PyTorch 2.8 dropped CUDA 11.x support, causing install_torch() to fail
with HTTP 403 when downloading libtorch cu117 binaries. Upgrade the full
GPU stack: CUDA 12.8.1, cuDNN 9.8, NCCL 2.25, TF 2.18 (CUDA 12.x compat,
GPU bundled since TF 2.15).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The rstats base image is Ubuntu 22.04 (Jammy, libc6 2.35). The NCCL
packages from the ubuntu2404 CUDA repo require libc6 >= 2.38, causing
apt dependency failures. Use the ubuntu2204 CUDA repo instead.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…yout

The nvidia/cuda 12.x images use cuda-ubuntu2204-x86_64.list (not
cuda.list) and store the GPG key at /usr/share/keyrings/ (not
/etc/apt/trusted.gpg).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@google-cla

google-cla Bot commented Apr 16, 2026

Copy link
Copy Markdown

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant