32 lines
1.1 KiB
Docker
32 lines
1.1 KiB
Docker
# llama-server für Qwen3.6 Vision — AMD Radeon AI Pro R9700, Vulkan-Backend
|
|
#
|
|
# Muss aus aktuellem Source gebaut werden (Qwen3.6-Rope-Änderung, s. Dockerfile.cuda).
|
|
# Vulkan statt ROCm — hat sich bei Vision als stabiler erwiesen.
|
|
|
|
FROM ubuntu:24.04 AS build
|
|
|
|
RUN apt-get update && apt-get install -y --no-install-recommends \
|
|
git cmake build-essential libcurl4-openssl-dev ca-certificates \
|
|
libvulkan-dev glslc \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
RUN git clone --depth 1 https://github.com/ggml-org/llama.cpp /src
|
|
|
|
RUN cmake /src -B /build \
|
|
-DGGML_VULKAN=ON \
|
|
-DBUILD_SHARED_LIBS=OFF \
|
|
-DLLAMA_CURL=ON \
|
|
&& cmake --build /build --config Release -j --target llama-server
|
|
|
|
FROM ubuntu:24.04
|
|
|
|
# mesa-vulkan-drivers = RADV-Treiber im Container (GPU via /dev/dri durchgereicht)
|
|
RUN apt-get update && apt-get install -y --no-install-recommends \
|
|
libvulkan1 mesa-vulkan-drivers vulkan-tools \
|
|
libcurl4 libgomp1 curl ca-certificates \
|
|
&& rm -rf /var/lib/apt/lists/*
|
|
|
|
COPY --from=build /build/bin/llama-server /usr/local/bin/llama-server
|
|
|
|
EXPOSE 8000
|
|
ENTRYPOINT ["llama-server"] |