Install diffusers

.github/workflows/release.yml

@@ -87,6 +87,12 @@ jobs:
             echo "docker/model-runner:latest-sglang-cuda" >> "$GITHUB_OUTPUT"
           fi
           echo 'EOF' >> "$GITHUB_OUTPUT"
+          echo "diffusers<<EOF" >> "$GITHUB_OUTPUT"
+          echo "docker/model-runner:${{ inputs.releaseTag }}-diffusers" >> "$GITHUB_OUTPUT"
+          if [ "${{ inputs.pushLatest }}" == "true" ]; then
+            echo "docker/model-runner:latest-diffusers" >> "$GITHUB_OUTPUT"
+          fi
+          echo 'EOF' >> "$GITHUB_OUTPUT"
           echo "rocm<<EOF" >> "$GITHUB_OUTPUT"
           echo "docker/model-runner:${{ inputs.releaseTag }}-rocm" >> "$GITHUB_OUTPUT"
           if [ "${{ inputs.pushLatest }}" == "true" ]; then
@@ -182,6 +188,19 @@ jobs:
           provenance: mode=max
           tags: ${{ steps.tags.outputs.sglang-cuda }}
 
+      - name: Build Diffusers image
+        uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
+        with:
+          file: Dockerfile
+          target: final-diffusers
+          platforms: linux/amd64, linux/arm64
+          build-args: |
+            "LLAMA_SERVER_VERSION=${{ inputs.llamaServerVersion }}"
+          push: true
+          sbom: true
+          provenance: mode=max
+          tags: ${{ steps.tags.outputs.diffusers }}
+
       - name: Build ROCm image
         uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83
         with:

cmd/cli/commands/install-runner.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"os"
+	"strings"
 	"time"
 
 	"github.com/docker/docker/api/types/container"
@@ -13,6 +14,7 @@ import (
 	gpupkg "github.com/docker/model-runner/cmd/cli/pkg/gpu"
 	"github.com/docker/model-runner/cmd/cli/pkg/standalone"
 	"github.com/docker/model-runner/cmd/cli/pkg/types"
+	"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/vllm"
 	"github.com/spf13/cobra"
@@ -26,7 +28,7 @@ const (
 	// installation will try to reach the model runner while waiting for it to
 	// be ready.
 	installWaitRetryInterval = 500 * time.Millisecond
-	backendUsage             = "Specify backend (" + llamacpp.Name + "|" + vllm.Name + "). Default: " + llamacpp.Name
+	backendUsage             = "Specify backend (" + llamacpp.Name + "|" + vllm.Name + "|" + diffusers.Name + "). Default: " + llamacpp.Name
 )
 
 // waitForStandaloneRunnerAfterInstall waits for a standalone model runner
@@ -322,8 +324,18 @@ func runInstallOrStart(cmd *cobra.Command, opts runnerOptions, debug bool) error
 	}
 
 	// Validate backend selection
-	if opts.backend != "" && opts.backend != llamacpp.Name && opts.backend != vllm.Name {
-		return fmt.Errorf("unknown backend: %q (supported: %s, %s)", opts.backend, llamacpp.Name, vllm.Name)
+	validBackends := []string{llamacpp.Name, vllm.Name, diffusers.Name}
+	if opts.backend != "" {
+		isValid := false
+		for _, valid := range validBackends {
+			if opts.backend == valid {
+				isValid = true
+				break
+			}
+		}
+		if !isValid {
+			return fmt.Errorf("unknown backend: %q (supported: %s)", opts.backend, strings.Join(validBackends, ", "))
+		}
 	}
 
 	// Validate backend-GPU compatibility

cmd/cli/docs/reference/docker_model_install-runner.yaml

@@ -8,7 +8,7 @@ plink: docker_model.yaml
 options:
     - option: backend
       value_type: string
-      description: 'Specify backend (llama.cpp|vllm). Default: llama.cpp'
+      description: 'Specify backend (llama.cpp|vllm|diffusers). Default: llama.cpp'
       deprecated: false
       hidden: false
       experimental: false

cmd/cli/docs/reference/docker_model_reinstall-runner.yaml

@@ -8,7 +8,7 @@ plink: docker_model.yaml
 options:
     - option: backend
       value_type: string
-      description: 'Specify backend (llama.cpp|vllm). Default: llama.cpp'
+      description: 'Specify backend (llama.cpp|vllm|diffusers). Default: llama.cpp'
       deprecated: false
       hidden: false
       experimental: false

cmd/cli/docs/reference/docker_model_start-runner.yaml

@@ -10,7 +10,7 @@ plink: docker_model.yaml
 options:
     - option: backend
       value_type: string
-      description: 'Specify backend (llama.cpp|vllm). Default: llama.cpp'
+      description: 'Specify backend (llama.cpp|vllm|diffusers). Default: llama.cpp'
       deprecated: false
       hidden: false
       experimental: false

cmd/cli/docs/reference/model_install-runner.md

@@ -7,7 +7,7 @@ Install Docker Model Runner (Docker Engine only)
 
 | Name             | Type     | Default     | Description                                                                                            |
 |:-----------------|:---------|:------------|:-------------------------------------------------------------------------------------------------------|
-| `--backend`      | `string` |             | Specify backend (llama.cpp\|vllm). Default: llama.cpp                                                  |
+| `--backend`      | `string` |             | Specify backend (llama.cpp\|vllm\|diffusers). Default: llama.cpp                                       |
 | `--debug`        | `bool`   |             | Enable debug logging                                                                                   |
 | `--do-not-track` | `bool`   |             | Do not track models usage in Docker Model Runner                                                       |
 | `--gpu`          | `string` | `auto`      | Specify GPU support (none\|auto\|cuda\|rocm\|musa\|cann)                                               |

cmd/cli/docs/reference/model_reinstall-runner.md

@@ -7,7 +7,7 @@ Reinstall Docker Model Runner (Docker Engine only)
 
 | Name             | Type     | Default     | Description                                                                                            |
 |:-----------------|:---------|:------------|:-------------------------------------------------------------------------------------------------------|
-| `--backend`      | `string` |             | Specify backend (llama.cpp\|vllm). Default: llama.cpp                                                  |
+| `--backend`      | `string` |             | Specify backend (llama.cpp\|vllm\|diffusers). Default: llama.cpp                                       |
 | `--debug`        | `bool`   |             | Enable debug logging                                                                                   |
 | `--do-not-track` | `bool`   |             | Do not track models usage in Docker Model Runner                                                       |
 | `--gpu`          | `string` | `auto`      | Specify GPU support (none\|auto\|cuda\|rocm\|musa\|cann)                                               |

cmd/cli/docs/reference/model_start-runner.md

@@ -7,7 +7,7 @@ Start Docker Model Runner (Docker Engine only)
 
 | Name             | Type     | Default     | Description                                                                                            |
 |:-----------------|:---------|:------------|:-------------------------------------------------------------------------------------------------------|
-| `--backend`      | `string` |             | Specify backend (llama.cpp\|vllm). Default: llama.cpp                                                  |
+| `--backend`      | `string` |             | Specify backend (llama.cpp\|vllm\|diffusers). Default: llama.cpp                                       |
 | `--debug`        | `bool`   |             | Enable debug logging                                                                                   |
 | `--do-not-track` | `bool`   |             | Do not track models usage in Docker Model Runner                                                       |
 | `--gpu`          | `string` | `auto`      | Specify GPU support (none\|auto\|cuda\|rocm\|musa\|cann)                                               |

cmd/cli/pkg/standalone/controller_image.go

@@ -4,6 +4,7 @@ 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"
 )
 
@@ -32,6 +33,10 @@ 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: