feat: add vLLM ROCm Docker variant for AMD GPUs

Adds a new Docker build target (final-vllm-rocm) that builds upstream
vLLM from source on top of rocm/vllm-dev:base (AMD's pre-built ROCm dev
image containing PyTorch ROCm, Triton, flash-attention, and the ROCm SDK).

vLLM is checked out at the tagged release matching VLLM_VERSION — no
fork, no custom wheels. The runtime stage mirrors the /opt/vllm-env
layout that the existing vLLM backend expects via symlinks, so no Go
code changes are needed.

Unlike the CUDA vllm variant, the ROCm image does not include llama.cpp:
the base images are incompatible (different ROCm runtime versions), so
the ROCm vllm image is intended as a vLLM-only artifact.

Fixes #412

Author: dmedovich

Dockerfile

@@ -155,3 +155,66 @@ COPY --from=builder /app/model-runner /app/model-runner
 FROM sglang AS final-sglang
 # Copy the built binary from builder-sglang (without vLLM)
 COPY --from=builder-sglang /app/model-runner /app/model-runner
+
+# --- vLLM ROCm: builder stage ---
+# Builds upstream vLLM from source on AMD's pre-built ROCm dev image, which
+# already contains PyTorch ROCm, Triton, flash-attention, and the ROCm SDK
+# (see https://hub.docker.com/r/rocm/vllm-dev). vLLM is checked out at the
+# tagged release matching VLLM_VERSION — no fork, no custom wheels.
+FROM rocm/vllm-dev:base AS vllm-rocm-builder
+
+ARG VLLM_VERSION=0.19.1
+# Target GPU architectures officially supported by vLLM ROCm:
+# gfx90a (MI200), gfx942 (MI300), gfx1100/1101 (RDNA3 7900/7800).
+ARG PYTORCH_ROCM_ARCH="gfx90a;gfx942;gfx1100;gfx1101"
+ENV PYTORCH_ROCM_ARCH=${PYTORCH_ROCM_ARCH}
+
+RUN git clone --depth 1 --branch v${VLLM_VERSION} \
+    https://github.com/vllm-project/vllm.git /vllm-src
+
+WORKDIR /vllm-src
+RUN python3 -m pip install --no-cache-dir -r requirements/rocm.txt \
+    && python3 setup.py bdist_wheel --dist-dir=/wheels
+
+# --- vLLM ROCm: runtime stage ---
+# Mirrors the /opt/vllm-env layout that pkg/inference/backends/vllm/vllm.go
+# expects (binary at /opt/vllm-env/bin/vllm, version file at
+# /opt/vllm-env/version). Symlinks are used instead of a real venv because
+# rocm/vllm-dev:base installs Python dependencies system-wide and recreating
+# a venv would break the PyTorch ROCm / Triton ROCm wiring.
+#
+# Note: unlike the CUDA vllm stage, this image does NOT include llama.cpp.
+# The base image is incompatible (different ROCm runtime versions), and the
+# rocm vllm image is intended as a vLLM-only artifact.
+FROM rocm/vllm-dev:base AS vllm-rocm
+
+COPY --from=vllm-rocm-builder /wheels/*.whl /tmp/
+RUN python3 -m pip install --no-cache-dir /tmp/*.whl && rm /tmp/*.whl
+
+RUN groupadd --system modelrunner \
+    && useradd --system --gid modelrunner -G video \
+        --create-home --home-dir /home/modelrunner modelrunner
+
+RUN mkdir -p /opt/vllm-env/bin \
+    && ln -s "$(command -v vllm)" /opt/vllm-env/bin/vllm \
+    && python3 -c "import vllm; print(vllm.__version__)" > /opt/vllm-env/version \
+    && chown -R modelrunner:modelrunner /opt/vllm-env
+
+RUN mkdir -p /var/run/model-runner /models /app \
+    && chown -R modelrunner:modelrunner /var/run/model-runner /app /models \
+    && chmod -R 755 /models
+
+USER modelrunner
+
+ENV MODEL_RUNNER_SOCK=/var/run/model-runner/model-runner.sock
+ENV MODEL_RUNNER_PORT=12434
+ENV HOME=/home/modelrunner
+ENV MODELS_PATH=/models
+
+LABEL com.docker.desktop.service="model-runner"
+
+ENTRYPOINT ["/app/model-runner"]
+
+FROM vllm-rocm AS final-vllm-rocm
+# Copy the built binary from builder
+COPY --from=builder /app/model-runner /app/model-runner

Makefile

@@ -9,6 +9,7 @@ LLAMA_UPSTREAM_IMAGE ?= $(shell \
 	"$(LLAMA_SERVER_VERSION)" "$(LLAMA_SERVER_VARIANT)")
 DOCKER_IMAGE := docker/model-runner:latest
 DOCKER_IMAGE_VLLM := docker/model-runner:latest-vllm-cuda
+DOCKER_IMAGE_VLLM_ROCM := docker/model-runner:latest-vllm-rocm
 DOCKER_IMAGE_SGLANG := docker/model-runner:latest-sglang
 DOCKER_IMAGE_MUSA := docker/model-runner:latest-musa
 DOCKER_IMAGE_OPENVINO := docker/model-runner:latest-openvino
@@ -43,7 +44,7 @@ DOCKER_BUILD_COMMON_ARGS = \
 .PHONY: build build-cli build-dmr build-llamacpp install-cli run clean test integration-tests e2e
 .PHONY: validate validate-versions validate-all lint help
 .PHONY: docker-build docker-build-multiplatform docker-run docker-run-impl
-.PHONY: docker-build-vllm docker-run-vllm docker-build-sglang docker-run-sglang
+.PHONY: docker-build-vllm docker-run-vllm docker-build-vllm-rocm docker-run-vllm-rocm docker-build-sglang docker-run-sglang
 .PHONY: docker-build-musa docker-run-musa docker-build-openvino docker-run-openvino
 .PHONY: test-docker-ce-installation
 .PHONY: vllm-metal-build vllm-metal-install vllm-metal-dev vllm-metal-clean
@@ -194,6 +195,23 @@ docker-build-vllm:
 docker-run-vllm: docker-build-vllm
 	@$(MAKE) -s docker-run-impl DOCKER_IMAGE=$(DOCKER_IMAGE_VLLM)
 
+# Build vLLM Docker image with ROCm (AMD GPU) support.
+# Builds upstream vLLM from source on top of rocm/vllm-dev:base — this is a
+# vLLM-only image (no llama.cpp), unlike the CUDA variant. Build is heavy:
+# expect 30-60 min and ~12-15 GB final image size.
+# LLAMA_SERVER_VARIANT is not consumed by the Dockerfile stages here, but
+# setting it to "rocm" restricts DOCKER_BUILD_PLATFORMS to linux/amd64
+# (vLLM ROCm has no aarch64 support).
+docker-build-vllm-rocm:
+	@$(MAKE) docker-build \
+		DOCKER_TARGET=final-vllm-rocm \
+		DOCKER_IMAGE=$(DOCKER_IMAGE_VLLM_ROCM) \
+		LLAMA_SERVER_VARIANT=rocm
+
+# Run vLLM ROCm Docker container with TCP port access and mounted model storage
+docker-run-vllm-rocm: docker-build-vllm-rocm
+	@$(MAKE) -s docker-run-impl DOCKER_IMAGE=$(DOCKER_IMAGE_VLLM_ROCM)
+
 # Build SGLang Docker image
 docker-build-sglang:
 	@$(MAKE) docker-build \
@@ -402,8 +420,10 @@ help:
 	@echo "  docker-build			- Build Docker image for current platform"
 	@echo "  docker-build-multiplatform	- Build Docker image for multiple platforms"
 	@echo "  docker-run			- Run in Docker container with TCP port access and mounted model storage"
-	@echo "  docker-build-vllm		- Build vLLM Docker image"
-	@echo "  docker-run-vllm		- Run vLLM Docker container"
+	@echo "  docker-build-vllm		- Build vLLM Docker image (CUDA)"
+	@echo "  docker-run-vllm		- Run vLLM Docker container (CUDA)"
+	@echo "  docker-build-vllm-rocm	- Build vLLM Docker image (ROCm / AMD GPU, source build)"
+	@echo "  docker-run-vllm-rocm		- Run vLLM Docker container (ROCm / AMD GPU)"
 	@echo "  docker-build-sglang		- Build SGLang Docker image"
 	@echo "  docker-run-sglang		- Run SGLang Docker container"
 	@echo "  docker-build-musa		- Build MUSA Docker image"