Debian Patches
Status for ggml-cuda/0.16.0-1
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| cmake-Add-option-to-install-tests.patch | cmake: Add option to install tests | Christian Kastner <ckk@debian.org> | not-needed | 2025-05-18 | ||
| Downgrade-ROCm-dependency-to-5.5.patch | Downgrade ROCm dependency to 5.5 The downgrade is necessary to have this package buildable in trixie and in current unstable. This requires a fixed rocm-hipamd 5.7.1, see http://bugs.debian.org/1116585 Furthermore, this patch addresses an issue gfx110n cards in 5.5, and with hipblas < 5.6. |
Christian Kastner <ckk@debian.org> | not-needed | https://github.com/ggml-org/llama.cpp/pull/15296 | 2026-02-17 | |
| Use-only-CPU-backend-in-build-time-tests.patch | Use only CPU backend in build-time tests | Christian Kastner <ckk@debian.org> | not-needed | 2026-02-03 | ||
| Take-over-some-tests-from-llama.cpp.patch | Take over some tests from llama.cpp Some pure-ggml tests are kept in the llama.cpp source because they are needed there for rapid iteration. They will eventually be moved to the ggml source. Debian implements this move earlier so that we can drop the embedded ggml source from llama.cpp. See also github.com/ggml-org/llama.cpp/pull/25179 |
Christian Kastner <ckk@debian.org> | not-needed | 2026-07-07 |
All known versions for source package 'ggml-cuda'
- 0.16.0-1 (forky, sid)