Skip to content

Fix llama.cpp ROCm build: install rocBLAS/hipBLAS dev libs - #14

Merged
Aaronontheweb merged 1 commit into
masterfrom
fix/amd-rocm-hipblas-dev
Aug 14, 2026
Merged

Aaronontheweb merged 1 commit into
masterfrom
fix/amd-rocm-hipblas-dev

Conversation

@Aaronontheweb

Copy link
Copy Markdown
Member

Second candidate build failed at cmake configure:

CMake Error at ggml/src/ggml-hip/CMakeLists.txt:47 (find_package):
  Could not find a package configuration file provided by "hipblas"

The rocm/dev-ubuntu-24.04 base ships the HIP compiler but not rocBLAS/hipBLAS. This installs rocblas-dev/hipblas-dev (CMake configs + runtime libs), adds CMAKE_PREFIX_PATH=/opt/rocm, and builds in a single stage so the runtime libraries ship with llama-server. Lock data updated. Re-dispatch the candidate build after merge.

Second candidate build failed at cmake configure: ggml-hip's
find_package(hipblas) could not find hipblas-config.cmake because the ROCm
dev base ships the HIP compiler but not the rocBLAS/hipBLAS packages.

Install rocblas-dev + hipblas-dev (they provide the CMake configs and pull
the runtime libraries llama-server links), add CMAKE_PREFIX_PATH=/opt/rocm,
and collapse to a single stage so those runtime libraries ship in the final
image. Update the dependency lock to record the added apt packages.
@Aaronontheweb
Aaronontheweb merged commit b5bd007 into master Aug 14, 2026
1 check passed
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