Misc. bug: ROCm images cannot be found

[ExposedCat]

Name and Version

Operating systems

Linux

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

llama-server

Command line

`podman pull ghcr.io/ggml-org/llama.cpp:server-rocm`

Problem description & steps to reproduce

Docs mention server-rocm docker image, but it seems to be missing along with any other ROCm images:

$ skopeo list-tags docker://ghcr.io/ggml-org/llama.cpp

{
    "Repository": "ghcr.io/ggml-org/llama.cpp",
    "Tags": [
        "full-vulkan",
        "full-musa",
        "full-musa-b4721",
        "light-musa",
        "light-musa-b4721",
        "server-musa",
        "server-musa-b4721",
        "full-cuda",
        "full-cuda-b4721",
        "light-cuda",
        "light-cuda-b4721",
        "server-cuda",
        "server-cuda-b4721",
        "full-vulkan-b4726",
        "full-intel",
        "full-intel-b4726",
        "light-vulkan",
        "light-vulkan-b4726",
        "server-vulkan",
        "server-vulkan-b4726",
        "light-intel",
        "light-intel-b4726",
        "server-intel",
        "server-intel-b4726",
        "full-musa-b4726",
        "light-musa-b4726",
        "server-musa-b4726",
        "full-cuda-b4726",
        "light-cuda-b4726",
        "server-cuda-b4726",
        "full-intel-b4729",
        "full-vulkan-b4729",
        "light-vulkan-b4729",
        "server-vulkan-b4729",
        "light-intel-b4729",
        "server-intel-b4729",
        "full-musa-b4729",
        "light-musa-b4729",
        "server-musa-b4729",
        "full-cuda-b4729",
        "light-cuda-b4729",
        "server-cuda-b4729",
        "full",
        "full-b4729",
        "light",
        "light-b4729",
        "server",
        "server-b4729"
    ]
}

First Bad Commit

No response

Relevant log output

[aviallon]

I confirm the bug. They are not built.

[ngxson]

See:

llama.cpp/.github/workflows/docker.yml

Lines 44 to 45 in 2eea03d

	# Note: the rocm images are failing due to a compiler error and are disabled until this is fixed to allow the workflow to complete
	#- {tag: "rocm", dockerfile: ".devops/rocm.Dockerfile", platforms: "linux/amd64,linux/arm64", full: true, light: true, server: true, freediskspace: true }

[aviallon]

@ngxson just tried to build llama.cpp with the latest Ubuntu ROCm dev image, and it works, so I guess it could be re-enabled.

[github-actions]

This issue was closed because it has been inactive for 14 days since being marked as stale.

[davispuh]

This is still issue, I'm looking for rocm Docker images.

[grigio]

Up, why is this issue closed ?

[luckydevil13]

up