ai/ollama/Dockerfile

@@ -2,17 +2,17 @@
 #
 # Custom ollama image with ROCm 6.1 + gfx906 (MI50) support.
 # The default ollama/rocm image ships ROCm 7.2 which dropped gfx906 support.
-# This builds ollama from source targeting AMDGPU_TARGETS=gfx906.
+# This builds ollama and its llama.cpp runner from source, targeting gfx906.
 #
-# Build: docker build -t ollama/ollama:rocm-gfx906 .
+# Build: docker build -t ollama/ollama:rocm-gfx906 ./ai/ollama
 
 FROM rocm/dev-ubuntu-22.04:6.1.2-complete AS builder
 
 RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y \
-    git golang cmake build-essential pkg-config \
+    git golang-go cmake build-essential pkg-config \
     && rm -rf /var/lib/apt/lists/*
 
-ARG OLLAMA_VERSION=v0.13.5
+ARG OLLAMA_VERSION=v0.23.2
 RUN git clone --depth 1 --branch ${OLLAMA_VERSION} https://github.com/ollama/ollama.git /build
 WORKDIR /build
 
@@ -20,7 +20,7 @@ ENV HIP_PATH=/opt/rocm
 ENV ROCM_PATH=/opt/rocm
 ENV PATH=/opt/rocm/bin:/opt/rocm/hip/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
 RUN cd llama.cpp && \
-    mkdir build && cd build && \
+    mkdir -p build && cd build && \
     cmake .. \
       -DLLAMA_HIPBLAS=ON \
       -DCMAKE_C_COMPILER=clang \
@@ -52,6 +52,9 @@ RUN ldconfig
 
 ENV LD_LIBRARY_PATH=/opt/rocm/lib:/usr/lib/ollama
 ENV HSA_OVERRIDE_GFX_VERSION=9.0.6
+ENV HCC_AMDGPU_TARGET=gfx906
+ENV HSA_ENABLE_SDMA=0
 
 EXPOSE 11434
+ENTRYPOINT ["/bin/ollama"]
 CMD ["serve"]