Name and Version

I just run:

$  docker pull ghcr.io/ggml-org/llama.cpp:server-cuda
...
Digest: sha256:0d60155f9cbd5118d02568d90f505638259d85f6f1cc4ac98d0f1002001e1f7a

Trying to run that container is now failing with:

$  docker run --gpus all ghcr.io/ggml-org/llama.cpp:server-cuda
docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running prestart hook #0: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: requirement error: unsatisfied condition: cuda>=12.9, please update your driver to a newer version, or use an earlier cuda container: unknown

The version of yesterday was running fine on my system (nvidia Driver Version: 570.211.01 CUDA Version: 12.8)

It was mentioned nowhere that the minimum required CUDA version was raised to 12.9
And looking at cloud providers like vast I see many instances running with CUDA less than 12.9, so those older versions are still highly relevant and a raise of the minimum requirements should be considered only very carefully.

Operating systems

Linux

Which llama.cpp modules do you know to be affected?

llama-server

Command line

`docker run --gpus all ghcr.io/ggml-org/llama.cpp:server-cuda`

Problem description & steps to reproduce

It doesn't start and shows a minimum CUDA version >= 12.9 error

First Bad Commit

No response

Relevant log output

No response