Remove unused final-diffusers Docker image variant and fix duplicate backend registration

Dockerfile

@@ -147,59 +147,6 @@ RUN curl -LsSf https://astral.sh/uv/install.sh | sh \
 
 RUN /opt/sglang-env/bin/python -c "import sglang; print(sglang.__version__)" > /opt/sglang-env/version
 
-# --- Diffusers variant ---
-FROM llamacpp AS diffusers
-
-# Python package versions for reproducible builds
-ARG DIFFUSERS_VERSION=0.36.0
-ARG TORCH_VERSION=2.9.1
-ARG TRANSFORMERS_VERSION=4.57.5
-ARG ACCELERATE_VERSION=1.3.0
-ARG SAFETENSORS_VERSION=0.5.2
-ARG HUGGINGFACE_HUB_VERSION=0.34.0
-ARG BITSANDBYTES_VERSION=0.49.1
-ARG FASTAPI_VERSION=0.115.12
-ARG UVICORN_VERSION=0.34.1
-ARG PILLOW_VERSION=11.2.1
-
-USER root
-
-RUN apt update && apt install -y \
-    python3 python3-venv python3-dev \
-    curl ca-certificates build-essential \
-    && rm -rf /var/lib/apt/lists/*
-
-RUN mkdir -p /opt/diffusers-env && chown -R modelrunner:modelrunner /opt/diffusers-env
-
-USER modelrunner
-
-# Install uv and diffusers as modelrunner user
-RUN curl -LsSf https://astral.sh/uv/install.sh | sh \
-    && ~/.local/bin/uv venv --python /usr/bin/python3 /opt/diffusers-env \
-    && ~/.local/bin/uv pip install --python /opt/diffusers-env/bin/python \
-    "diffusers==${DIFFUSERS_VERSION}" \
-    "torch==${TORCH_VERSION}" \
-    "transformers==${TRANSFORMERS_VERSION}" \
-    "accelerate==${ACCELERATE_VERSION}" \
-    "safetensors==${SAFETENSORS_VERSION}" \
-    "huggingface_hub==${HUGGINGFACE_HUB_VERSION}" \
-    "bitsandbytes==${BITSANDBYTES_VERSION}" \
-    "fastapi==${FASTAPI_VERSION}" \
-    "uvicorn[standard]==${UVICORN_VERSION}" \
-    "pillow==${PILLOW_VERSION}"
-
-# Copy Python server code
-USER root
-COPY python/diffusers_server /tmp/diffusers_server/
-RUN PYTHON_SITE_PACKAGES=$(/opt/diffusers-env/bin/python -c "import site; print(site.getsitepackages()[0])") && \
-    mkdir -p "$PYTHON_SITE_PACKAGES/diffusers_server" && \
-    cp -r /tmp/diffusers_server/* "$PYTHON_SITE_PACKAGES/diffusers_server/" && \
-    chown -R modelrunner:modelrunner "$PYTHON_SITE_PACKAGES/diffusers_server/" && \
-    rm -rf /tmp/diffusers_server
-USER modelrunner
-
-RUN /opt/diffusers-env/bin/python -c "import diffusers; print(diffusers.__version__)" > /opt/diffusers-env/version
-
 FROM llamacpp AS final-llamacpp
 # Copy the built binary from builder
 COPY --from=builder /app/model-runner /app/model-runner
@@ -211,7 +158,3 @@ 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
-
-FROM diffusers AS final-diffusers
-# Copy the built binary from builder (with diffusers support)
-COPY --from=builder /app/model-runner /app/model-runner

Makefile

@@ -9,7 +9,6 @@ VLLM_VERSION ?= 0.17.0
 DOCKER_IMAGE := docker/model-runner:latest
 DOCKER_IMAGE_VLLM := docker/model-runner:latest-vllm-cuda
 DOCKER_IMAGE_SGLANG := docker/model-runner:latest-sglang
-DOCKER_IMAGE_DIFFUSERS := docker/model-runner:latest-diffusers
 DOCKER_TARGET ?= final-llamacpp
 PORT := 8080
 LLAMA_ARGS ?=
@@ -31,7 +30,6 @@ BUILD_DMR ?= 1
 .PHONY: validate 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-diffusers docker-run-diffusers
 .PHONY: test-docker-ce-installation
 .PHONY: vllm-metal-build vllm-metal-install vllm-metal-dev vllm-metal-clean
 .PHONY: diffusers-build diffusers-install diffusers-dev diffusers-clean
@@ -151,16 +149,6 @@ docker-build-sglang:
 docker-run-sglang: docker-build-sglang
 	@$(MAKE) -s docker-run-impl DOCKER_IMAGE=$(DOCKER_IMAGE_SGLANG)
 
-# Build Diffusers Docker image
-docker-build-diffusers:
-	@$(MAKE) docker-build \
-		DOCKER_TARGET=final-diffusers \
-		DOCKER_IMAGE=$(DOCKER_IMAGE_DIFFUSERS)
-
-# Run Diffusers Docker container with TCP port access and mounted model storage
-docker-run-diffusers: docker-build-diffusers
-	@$(MAKE) -s docker-run-impl DOCKER_IMAGE=$(DOCKER_IMAGE_DIFFUSERS)
-
 # Common implementation for running Docker container
 docker-run-impl:
 	@echo ""
@@ -338,8 +326,6 @@ help:
 	@echo "  docker-run-vllm		- Run vLLM Docker container"
 	@echo "  docker-build-sglang		- Build SGLang Docker image"
 	@echo "  docker-run-sglang		- Run SGLang Docker container"
-	@echo "  docker-build-diffusers	- Build Diffusers Docker image"
-	@echo "  docker-run-diffusers		- Run Diffusers Docker container"
 	@echo "  vllm-metal-build		- Build vllm-metal tarball locally (macOS ARM64)"
 	@echo "  vllm-metal-install		- Install vllm-metal from local tarball"
 	@echo "  vllm-metal-dev		- Install vllm-metal from local source (editable)"

cmd/cli/pkg/standalone/controller_image.go

@@ -4,7 +4,6 @@ import (
 	"os"
 
 	gpupkg "github.com/docker/model-runner/cmd/cli/pkg/gpu"
-	"github.com/docker/model-runner/pkg/inference/backends/diffusers"
 	"github.com/docker/model-runner/pkg/inference/backends/vllm"
 )
 
@@ -33,10 +32,6 @@ func controllerImageVariant(detectedGPU gpupkg.GPUSupport, backend string) strin
 	if backend == vllm.Name {
 		return "vllm-cuda"
 	}
-	// If diffusers backend is requested, return diffusers variant
-	if backend == diffusers.Name {
-		return "diffusers"
-	}
 	// Default to llama.cpp backend behavior
 	switch detectedGPU {
 	case gpupkg.GPUSupportCUDA:

main.go

@@ -17,7 +17,6 @@ import (
 
 	"github.com/docker/model-runner/pkg/envconfig"
 	"github.com/docker/model-runner/pkg/inference"
-	"github.com/docker/model-runner/pkg/inference/backends/diffusers"
 	"github.com/docker/model-runner/pkg/inference/backends/llamacpp"
 	"github.com/docker/model-runner/pkg/inference/backends/sglang"
 	"github.com/docker/model-runner/pkg/inference/config"
@@ -132,9 +131,6 @@ func main() {
 			routing.BackendDef{Name: sglang.Name, Init: func(mm *models.Manager) (inference.Backend, error) {
 				return sglang.New(log, mm, log.With("component", sglang.Name), nil, sglangServerPath)
 			}},
-			routing.BackendDef{Name: diffusers.Name, Init: func(mm *models.Manager) (inference.Backend, error) {
-				return diffusers.New(log, mm, log.With("component", diffusers.Name), nil, diffusersServerPath)
-			}},
 		),
 		OnBackendError: func(name string, err error) {
 			log.Error("unable to initialize backend", "backend", name, "error", err)

scripts/docker-run.sh

@@ -3,7 +3,6 @@
 add_accelerators() {
   # Add NVIDIA GPU support for CUDA variants and GPU-accelerated backends
   if [[ "${DOCKER_IMAGE-}" == *"-cuda" ]] || \
-     [[ "${DOCKER_IMAGE-}" == *"-diffusers" ]] || \
      [[ "${DOCKER_IMAGE-}" == *"-sglang" ]]; then
       if docker info -f '{{range $k, $v := .Runtimes}}{{$k}}{{"\n"}}{{end}}' 2>/dev/null | grep -qx "nvidia"; then
         args+=("--gpus" "all" "--runtime=nvidia")