ci: add ROCm Dockerfile (#2094)

docker/Dockerfile.rocm

@@ -0,0 +1,50 @@
+FROM rocm/dev-ubuntu-24.04 AS build
+
+# sd-server embeds the web UI at build time, so the build image needs Node/pnpm.
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        build-essential git cmake ninja-build ccache ca-certificates curl gnupg && \
+    mkdir -p /etc/apt/keyrings && \
+    curl -fsSL https://deb.nodesource.com/gpgkey/nodesource-repo.gpg.key -o /tmp/nodesource-repo.gpg.key && \
+    gpg --dearmor -o /etc/apt/keyrings/nodesource.gpg /tmp/nodesource-repo.gpg.key && \
+    rm /tmp/nodesource-repo.gpg.key && \
+    echo "deb [signed-by=/etc/apt/keyrings/nodesource.gpg] https://deb.nodesource.com/node_20.x nodistro main" > /etc/apt/sources.list.d/nodesource.list && \
+    apt-get update && apt-get install -y --no-install-recommends nodejs && \
+    npm install -g pnpm@10.15.1 && \
+    apt-get clean && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /sd.cpp
+
+# ROCm's clang is not on PATH in Docker RUN shells (profile scripts are not loaded).
+ENV PATH="/opt/rocm/llvm/bin:$PATH"
+
+COPY . .
+
+ARG AMDGPU_TARGETS=""
+
+RUN cmake . -B ./build -G Ninja \
+    -DCMAKE_C_COMPILER=clang \
+    -DCMAKE_CXX_COMPILER=clang++ \
+    -DSD_HIPBLAS=ON \
+    -DGGML_NATIVE=OFF \
+    -DAMDGPU_TARGETS="${AMDGPU_TARGETS}" \
+    -DCMAKE_BUILD_TYPE=Release \
+    -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
+    -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+RUN cmake --build ./build --config Release --parallel
+
+FROM rocm/dev-ubuntu-24.04 AS runtime
+
+# ROCm clang links against its own OpenMP runtime in /opt/rocm/llvm/lib,
+# and the ROCm runtime libs (hipblas, rocblas, amdhip64) live in /opt/rocm/lib.
+ENV LD_LIBRARY_PATH="/opt/rocm/llvm/lib:/opt/rocm/lib:$LD_LIBRARY_PATH"
+
+RUN apt-get update && \
+    apt-get install --yes --no-install-recommends libgomp1 && \
+    apt-get clean && rm -rf /var/lib/apt/lists/*
+
+COPY --from=build /sd.cpp/build/bin /sd.cpp/bin
+RUN printf '#!/bin/sh\nexec /sd.cpp/bin/sd-cli "$@"\n' > /sd-cli && \
+    printf '#!/bin/sh\nexec /sd.cpp/bin/sd-server "$@"\n' > /sd-server && \
+    chmod +x /sd-cli /sd-server
+
+ENTRYPOINT [ "/sd-cli" ]

docs/docker.md

@@ -24,12 +24,25 @@ docker build -f docker/Dockerfile -t sd .
 
 ## Building variants using Docker
 
-Vulkan:
+### Vulkan:
 
 ```shell
 docker build -f docker/Dockerfile.vulkan -t sd .
 ```
 
+### ROCm (AMD GPU):
+
+Defining one or multiple targets with `--build-arg AMDGPU_TARGETS` is strongly recommended.
+Building the image will otherwise likely fail or build for a target that does not fit your
+hardware.
+
+```shell
+# Build for one target
+docker build -f docker/Dockerfile.rocm --build-arg "AMDGPU_TARGETS=gfx1100" -t sd .
+# Build for multiple targets
+docker build -f docker/Dockerfile.rocm --build-arg "AMDGPU_TARGETS=gfx1030;gfx1100" -t sd .
+```
+
 ## Run locally built image's CLI
 
 ```shell
@@ -37,3 +50,21 @@ docker run --rm -v /path/to/models:/models -v /path/to/output/:/output sd [args.
 # For example
 # docker run --rm -v ./models:/models -v ./build:/output sd -m /models/sd-v1-4.ckpt -p "a lovely cat" -v -o /output/output.png
 ```
+
+### ROCm (AMD GPU)
+
+The user that executes this command needs to be in the `video` group and have read and write
+access to `/dev/dri` and `/dev/kfd`.
+
+`--device` and `--group-add` allow the container access to the GPU.
+
+```shell
+docker run --rm --device=/dev/dri --device=/dev/kfd --group-add video \
+  -v /path/to/models:/models -v /path/to/output/:/output \
+  sd [args...]
+# For example
+# docker run --rm --device=/dev/dri --device=/dev/kfd --group-add video \
+#   -v ./models:/models -v ./build:/output \
+#   sd -m /models/sd-v1-4.ckpt -v -o /output/output.png \
+#   -p "a lovely cat"
+```