Publish context-aware inpainting v2 Core ML runtime

README.md

@@ -11,16 +11,12 @@ tags:
 
 # Clover Image Tiny Inpaint — Core ML
 
-This export is the SD 1.4-class target for the 9-channel Clover Image Tiny
-inpainting U-Net. The conversion is pinned to the Apple Stable Diffusion
-converter vendored under `.coreml-converter` and targets iOS 18.
+Compiled Core ML resources for the v2 SD 1.4-class Clover Image Tiny
+inpainting model. The optional download contains **1,671,581,989 bytes**:
+approximately **1,672 MB** (1.56 GiB). It is installed separately from Regular
+Clover.
 
-The downloadable resource manifest contains 1,670,353,234 bytes: approximately
-**1,670 MB** (1.56 GiB). It is installed separately from Regular Clover.
-
-## Resource contract
-
-The app-managed `Resources` directory contains:
+## Runtime resources
 
 ```text
 TextEncoder.mlmodelc
@@ -33,40 +29,38 @@ vocab.json
 merges.txt
 ```
 
-The U-Net input is `[1, 9, 64, 64]`, ordered as noisy latent, mask, and
-masked-image latent. The VAE encoder is required to create the final four
-channels on device. White mask pixels are regenerated and black pixels are
-preserved.
+The batch-one U-Net input is `[1, 9, 64, 64]`, ordered as noisy latent, mask,
+and masked-image latent. White mask pixels are regenerated; black pixels are
+preserved. The VAE encoder creates the masked-image latent locally.
 
-Example edits include replacing a masked object with a tiny greenhouse,
-removing a person while continuing the background, or adding a red kettle to a
-masked countertop. Use a grayscale mask where white means “regenerate” and
-black means “preserve.”
+Recommended settings are DPM-Solver++, 20 steps, CFG 6.0, and a 96-pixel
+context margin around focused mask crops. Composite the result through the
+exact mask to preserve all source pixels outside the edit.
 
-![Cat inpainted into a masked greenhouse doorway](https://huggingface.co/neonforestmist/Clover-Image-Tiny-Inpaint/resolve/main/examples/result-cat.png)
+![Context-aware cat inpaint](https://huggingface.co/neonforestmist/Clover-Image-Tiny-Inpaint/resolve/main/examples/cat-result.png)
 
-DPM-Solver++ at 20 steps is the recommended runtime configuration. Small masks
-benefit from a context crop before 512×512 inference, followed by exact-mask
-compositing back into the source. Regular Clover LoRAs target a different
-4-channel U-Net and cannot be loaded dynamically into this 9-channel export.
+## Validation
 
-## Convert and validate
+The converted nine-channel U-Net passed Apple conversion parity at 40.8 dB and
+the independent fixed-shape Core ML smoke test at **46.2 dB**, above the 35 dB
+release threshold. The manifest records every runtime file's size and SHA-256.
 
 ```bash
-coreml-tools/convert_inpaint.sh coreml-inpaint artifacts/clover-image-tiny-inpaint
-coreml-tools/smoke-test-inpaint.sh artifacts/clover-image-tiny-inpaint coreml-inpaint
+coreml-tools/convert_inpaint.sh \
+  coreml-inpaint-v2 artifacts/clover-image-tiny-inpaint-v2
+coreml-tools/smoke-test-inpaint.sh \
+  artifacts/clover-image-tiny-inpaint-v2 coreml-inpaint-v2
 .venv-coreml/bin/python coreml-tools/manifest_inpaint.py \
-  coreml-inpaint --model-dir artifacts/clover-image-tiny-inpaint
+  coreml-inpaint-release-v2 \
+  --model-dir artifacts/clover-image-tiny-inpaint-v2
 ```
 
-The validator checks the 9-channel Diffusers configuration, required compiled
-resources, the fixed Core ML input shape, finite output, and random-input
-PyTorch/Core ML parity. The all-accelerators path is used because the macOS
-CPU-only Core ML runtime can return NaNs for this MLProgram even though the
-Neural Engine-capable path is valid.
+The package targets iOS 18. The chunked batch-one U-Net lets the Swift runtime
+perform classifier-free guidance as two serial passes to reduce peak memory.
 
-The resulting `Resources` directory can be published as the companion model
-repo [`neonforestmist/Clover-Image-Tiny-Inpaint-CoreML`](https://huggingface.co/neonforestmist/Clover-Image-Tiny-Inpaint-CoreML).
+Regular Clover Image Tiny LoRAs target a four-channel U-Net and cannot be
+loaded dynamically into this nine-channel package. An inpainting-specific LoRA
+can be fused before conversion as a separate Core ML bundle.
 
 ## Citation
 

Resources/SafetyChecker.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8211508a22331e6868fa9d207c6532ae5fc11791e8ceac6bf6fa4c7dd89004e8
+size 243

Resources/SafetyChecker.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f57494e9bc01a089c73ba6b1b9554e1a3df155efbf7cf36223ba10ae494f0d58
+size 1549

Resources/SafetyChecker.mlmodelc/metadata.json

@@ -0,0 +1,130 @@
+[
+  {
+    "shortDescription" : "Stable Diffusion generates images conditioned on text and\/or other images as input through the diffusion process. Please refer to https:\/\/arxiv.org\/abs\/2112.10752 for details.",
+    "metadataOutputVersion" : "3.0",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 512 × 512 × 3)",
+        "shortDescription" : "Identical to the input `images`. If safety checker detected any sensitive content, the corresponding image is replaced with a blank image (zeros)",
+        "shape" : "[1, 512, 512, 3]",
+        "name" : "filtered_images",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1 × 1 × 1)",
+        "shortDescription" : "Indicates whether the safety checker model found any sensitive content in the given image",
+        "shape" : "[1, 1, 1, 1]",
+        "name" : "has_nsfw_concepts",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 17)",
+        "shortDescription" : "Concept scores are the scores before thresholding at zero yields the `has_nsfw_concepts` output. These scores can be used to tune the `adjustment` input",
+        "shape" : "[1, 17]",
+        "name" : "concept_scores",
+        "type" : "MultiArray"
+      }
+    ],
+    "version" : "clover-image-tiny-inpaint-v2",
+    "modelParameters" : [
+
+    ],
+    "author" : "Please refer to the Model Card available at huggingface.co\/clover-image-tiny-inpaint-v2",
+    "specificationVersion" : 9,
+    "storagePrecision" : "Float16",
+    "license" : "OpenRAIL (https:\/\/huggingface.co\/spaces\/CompVis\/stable-diffusion-license)",
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 4,
+      "Ios18.softmax" : 24,
+      "Ios18.mul" : 73,
+      "Ios18.matmul" : 48,
+      "Ios18.realDiv" : 1,
+      "Ios18.sigmoid" : 24,
+      "Ios18.equal" : 1,
+      "Ios18.greaterEqual" : 1,
+      "Select" : 2,
+      "Ios18.greater" : 4,
+      "Tile" : 4,
+      "Ios16.reduceSum" : 2,
+      "Ios18.add" : 52,
+      "Ios18.layerNorm" : 50,
+      "Ios18.scatterNd" : 1,
+      "Ios18.reshape" : 97,
+      "Shape" : 1,
+      "Ios18.maximum" : 1,
+      "Ios18.linear" : 147,
+      "Ios18.conv" : 1,
+      "Ios18.concat" : 2,
+      "Ios18.transpose" : 97,
+      "Ios18.cast" : 7,
+      "Ios18.sliceByIndex" : 2,
+      "NonZero" : 1,
+      "Ios16.reduceL2Norm" : 1
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "stateSchema" : [
+
+    ],
+    "isUpdatable" : "0",
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 3 × 224 × 224)",
+        "shortDescription" : "The normalized image input tensor resized to (224x224) in channels-first (BCHW) format",
+        "shape" : "[1, 3, 224, 224]",
+        "name" : "clip_input",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 512 × 512 × 3)",
+        "shortDescription" : "Output of the vae_decoder (512x512) in channels-last (BHWC) format",
+        "shape" : "[1, 512, 512, 3]",
+        "name" : "images",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1)",
+        "shortDescription" : "Bias added to the concept scores to trade off increased recall for reduce precision in the safety checker classifier",
+        "shape" : "[1]",
+        "name" : "adjustment",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.conversion_date" : "2026-08-11",
+      "com.github.apple.coremltools.source" : "torch==2.7.0",
+      "com.github.apple.coremltools.version" : "9.0",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    },
+    "generatedClassName" : "Stable_Diffusion_version_clover_image_tiny_inpaint_v2_safety_checker",
+    "method" : "predict"
+  }
+]

Resources/SafetyChecker.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffdfb1a4774feb816dd2ccc3e304d109013bc9591e188dcab840789583fdc1b4
+size 607990114

Resources/TextEncoder.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ba6d80c5e94f50f17a70a965c2f63d24fa4239672d0fb5dadcf888e69cafc0f
+size 243

Resources/TextEncoder.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:734dc3fdcdcc31367ffae6a58b1e4b0d713eb3f043b89811b7fcaec6df67a495
+size 952

Resources/TextEncoder.mlmodelc/metadata.json

@@ -0,0 +1,90 @@
+[
+  {
+    "shortDescription" : "Stable Diffusion generates images conditioned on text and\/or other images as input through the diffusion process. Please refer to https:\/\/arxiv.org\/abs\/2112.10752 for details.",
+    "metadataOutputVersion" : "3.0",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 77 × 768)",
+        "shortDescription" : "The token embeddings as encoded by the Transformer model",
+        "shape" : "[1, 77, 768]",
+        "name" : "last_hidden_state",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 768)",
+        "shortDescription" : "The version of the `last_hidden_state` output after pooling",
+        "shape" : "[1, 768]",
+        "name" : "pooled_outputs",
+        "type" : "MultiArray"
+      }
+    ],
+    "version" : "clover-image-tiny-inpaint-v2",
+    "modelParameters" : [
+
+    ],
+    "author" : "Please refer to the Model Card available at huggingface.co\/clover-image-tiny-inpaint-v2",
+    "specificationVersion" : 9,
+    "storagePrecision" : "Float16",
+    "license" : "OpenRAIL (https:\/\/huggingface.co\/spaces\/CompVis\/stable-diffusion-license)",
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.linear" : 72,
+      "Ios18.matmul" : 24,
+      "Select" : 1,
+      "Ios18.reduceArgmax" : 1,
+      "Ios18.gather" : 1,
+      "Ios18.sigmoid" : 12,
+      "Ios18.add" : 38,
+      "Stack" : 1,
+      "Ios18.softmax" : 12,
+      "Ios18.layerNorm" : 25,
+      "Ios18.transpose" : 48,
+      "Ios18.cast" : 4,
+      "Ios18.gatherNd" : 1,
+      "Ios18.reshape" : 48,
+      "Ios18.greaterEqual" : 1,
+      "Ios18.mul" : 36
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
+    "stateSchema" : [
+
+    ],
+    "isUpdatable" : "0",
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 77)",
+        "shortDescription" : "The token ids that represent the input text",
+        "shape" : "[1, 77]",
+        "name" : "input_ids",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.conversion_date" : "2026-08-11",
+      "com.github.apple.coremltools.source" : "torch==2.7.0",
+      "com.github.apple.coremltools.version" : "9.0",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    },
+    "generatedClassName" : "Stable_Diffusion_version_clover_image_tiny_inpaint_v2_text_encoder",
+    "method" : "predict"
+  }
+]

Resources/TextEncoder.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3e8699e18a386ba82633384e0a6f5c090983f1c66b24a26fad3554fc2a12ddb
+size 246145536

Resources/UnetChunk1.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ee3483e4486c1bdc21299013aa9d3a25fcafb3d064750294c822cf8d6708151
+size 243

Resources/UnetChunk1.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1ebbe33a5324d71a517efe9d5f5b8f9bbcb9bd2e991a57171840161b3a63430
+size 532

Resources/UnetChunk1.mlmodelc/metadata.json

@@ -0,0 +1,180 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Float16",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 320 × 64 × 64)",
+        "shortDescription" : "",
+        "shape" : "[1, 320, 64, 64]",
+        "name" : "input_7_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 640 × 32 × 32)",
+        "shortDescription" : "",
+        "shape" : "[1, 640, 32, 32]",
+        "name" : "input_63_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 10240 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 10240, 1, 256]",
+        "name" : "var_2337_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 640 × 16 × 16)",
+        "shortDescription" : "",
+        "shape" : "[1, 640, 16, 16]",
+        "name" : "input_65_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 320 × 32 × 32)",
+        "shortDescription" : "",
+        "shape" : "[1, 320, 32, 32]",
+        "name" : "input_37_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1280 × 16 × 16)",
+        "shortDescription" : "",
+        "shape" : "[1, 1280, 16, 16]",
+        "name" : "hidden_states_61_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 320 × 64 × 64)",
+        "shortDescription" : "",
+        "shape" : "[1, 320, 64, 64]",
+        "name" : "input_35_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1280 × 1 × 1)",
+        "shortDescription" : "",
+        "shape" : "[1, 1280, 1, 1]",
+        "name" : "input_15_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1280 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1280, 1, 256]",
+        "name" : "inputs_23_cast_fp16",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 3,
+      "Ios18.softmax" : 64,
+      "Ios18.cos" : 1,
+      "Ios18.mul" : 68,
+      "Ios18.batchNorm" : 12,
+      "Ios16.reduceMean" : 24,
+      "Ios18.realDiv" : 12,
+      "Ios18.sin" : 1,
+      "Split" : 3,
+      "Ios18.add" : 34,
+      "Ios18.layerNorm" : 12,
+      "Ios18.reshape" : 31,
+      "Ios18.conv" : 66,
+      "Ios18.concat" : 11,
+      "Ios16.einsum" : 128,
+      "Ios18.sub" : 12,
+      "Ios18.silu" : 10,
+      "Ios18.transpose" : 8,
+      "Ios18.sqrt" : 12,
+      "Ios18.gelu" : 3,
+      "Ios18.square" : 12,
+      "Ios18.cast" : 9,
+      "Ios18.sliceByIndex" : 194
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "isUpdatable" : "0",
+    "stateSchema" : [
+
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "userDefinedMetadata" : {
+
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 9 × 64 × 64)",
+        "shortDescription" : "",
+        "shape" : "[1, 9, 64, 64]",
+        "name" : "sample",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "timestep",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 768 × 1 × 77)",
+        "shortDescription" : "",
+        "shape" : "[1, 768, 1, 77]",
+        "name" : "encoder_hidden_states",
+        "type" : "MultiArray"
+      }
+    ],
+    "generatedClassName" : "Stable_Diffusion_version_clover_image_tiny_inpaint_v2_unet_chunk1",
+    "method" : "predict"
+  }
+]

Resources/UnetChunk1.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:091529bc9e2b2bdda0508614e223570f7a8a4fe288bc510ec844635ba2f01046
+size 324651584

Resources/UnetChunk2.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f7d7cba85c9edd629d133c0224baf71e989ad4bfe1da89f37706f1d6cab45ea
+size 243

Resources/UnetChunk2.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94ea7e153dab4f55ad8a40d12f9853422f8bdce240f5b608f3ae72ac3229fb35
+size 513

Resources/UnetChunk2.mlmodelc/metadata.json

@@ -0,0 +1,168 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Float16",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 4 × 64 × 64)",
+        "shortDescription" : "",
+        "shape" : "[1, 4, 64, 64]",
+        "name" : "noise_pred",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.softmax" : 80,
+      "Ios18.mul" : 86,
+      "UpsampleNearestNeighbor" : 2,
+      "Ios18.batchNorm" : 16,
+      "Ios16.reduceMean" : 32,
+      "Ios18.realDiv" : 16,
+      "Split" : 6,
+      "Ios18.add" : 48,
+      "Ios18.layerNorm" : 15,
+      "Ios18.reshape" : 43,
+      "Ios18.conv" : 85,
+      "Ios18.gelu" : 6,
+      "Ios18.concat" : 15,
+      "Ios18.sub" : 16,
+      "Ios16.einsum" : 160,
+      "Ios18.silu" : 11,
+      "Ios18.cast" : 10,
+      "Ios18.transpose" : 10,
+      "Ios18.square" : 16,
+      "Ios18.sqrt" : 16,
+      "Ios18.sliceByIndex" : 240
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "isUpdatable" : "0",
+    "stateSchema" : [
+
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "userDefinedMetadata" : {
+
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 768 × 1 × 77)",
+        "shortDescription" : "",
+        "shape" : "[1, 768, 1, 77]",
+        "name" : "encoder_hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 640 × 16 × 16)",
+        "shortDescription" : "",
+        "shape" : "[1, 640, 16, 16]",
+        "name" : "input_65_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1280 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1280, 1, 256]",
+        "name" : "inputs_23_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 320 × 64 × 64)",
+        "shortDescription" : "",
+        "shape" : "[1, 320, 64, 64]",
+        "name" : "input_35_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1280 × 16 × 16)",
+        "shortDescription" : "",
+        "shape" : "[1, 1280, 16, 16]",
+        "name" : "hidden_states_61_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 10240 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 10240, 1, 256]",
+        "name" : "var_2337_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 320 × 32 × 32)",
+        "shortDescription" : "",
+        "shape" : "[1, 320, 32, 32]",
+        "name" : "input_37_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 640 × 32 × 32)",
+        "shortDescription" : "",
+        "shape" : "[1, 640, 32, 32]",
+        "name" : "input_63_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 1280 × 1 × 1)",
+        "shortDescription" : "",
+        "shape" : "[1, 1280, 1, 1]",
+        "name" : "input_15_cast_fp16",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 320 × 64 × 64)",
+        "shortDescription" : "",
+        "shape" : "[1, 320, 64, 64]",
+        "name" : "input_7_cast_fp16",
+        "type" : "MultiArray"
+      }
+    ],
+    "generatedClassName" : "Stable_Diffusion_version_clover_image_tiny_inpaint_v2_unet_chunk2",
+    "method" : "predict"
+  }
+]

Resources/UnetChunk2.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04910ba6d4ca88b34dd1202c848aebfc33e339010d203573ebc4138a4e9f78ba
+size 322211328

Resources/VAEDecoder.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff20b000564aa02c66b7e254258a5c8ba7d34fdb99c54be3c85a42d8eea45a62
+size 243

Resources/VAEDecoder.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bb1f9ddd844d6c8e14c6e52a1e05d7c493e71091d788798d065141bbe3e131b
+size 879

Resources/VAEDecoder.mlmodelc/metadata.json

@@ -0,0 +1,80 @@
+[
+  {
+    "shortDescription" : "Stable Diffusion generates images conditioned on text and\/or other images as input through the diffusion process. Please refer to https:\/\/arxiv.org\/abs\/2112.10752 for details.",
+    "metadataOutputVersion" : "3.0",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 3 × 512 × 512)",
+        "shortDescription" : "Generated image normalized to range [-1, 1]",
+        "shape" : "[1, 3, 512, 512]",
+        "name" : "image",
+        "type" : "MultiArray"
+      }
+    ],
+    "version" : "clover-image-tiny-inpaint-v2",
+    "modelParameters" : [
+
+    ],
+    "author" : "Please refer to the Model Card available at huggingface.co\/clover-image-tiny-inpaint-v2",
+    "specificationVersion" : 9,
+    "storagePrecision" : "Float16",
+    "license" : "OpenRAIL (https:\/\/huggingface.co\/spaces\/CompVis\/stable-diffusion-license)",
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.square" : 30,
+      "Ios18.batchNorm" : 29,
+      "Ios18.conv" : 36,
+      "Ios18.sub" : 30,
+      "Ios18.linear" : 4,
+      "Ios18.scaledDotProductAttention" : 1,
+      "Ios18.add" : 46,
+      "Ios18.realDiv" : 30,
+      "Ios18.silu" : 29,
+      "Ios16.reduceMean" : 60,
+      "UpsampleNearestNeighbor" : 3,
+      "Ios18.transpose" : 6,
+      "Ios18.cast" : 1,
+      "Ios18.reshape" : 65,
+      "Ios18.sqrt" : 30,
+      "Ios18.mul" : 1
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "stateSchema" : [
+
+    ],
+    "isUpdatable" : "0",
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 4 × 64 × 64)",
+        "shortDescription" : "The denoised latent embeddings from the unet model after the last step of reverse diffusion",
+        "shape" : "[1, 4, 64, 64]",
+        "name" : "z",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.conversion_date" : "2026-08-11",
+      "com.github.apple.coremltools.source" : "torch==2.7.0",
+      "com.github.apple.coremltools.version" : "9.0",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    },
+    "generatedClassName" : "Stable_Diffusion_version_clover_image_tiny_inpaint_v2_vae_decoder",
+    "method" : "predict"
+  }
+]

Resources/VAEDecoder.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca45a093443d3a9719c5f7c61010315e94ffc8019eff27a0eeb52075a05ac752
+size 98993280

Resources/VAEEncoder.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba122c04ae4c8d28c8b8805049e22fb7dfeac3eada7bde68c2366a14e712d1d6
+size 243

Resources/VAEEncoder.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe1d98a943015c7ac3d29ef2d573b37610e62c97e5724d2f85c555960832629d
+size 883

Resources/VAEEncoder.mlmodelc/metadata.json

@@ -0,0 +1,80 @@
+[
+  {
+    "shortDescription" : "Stable Diffusion generates images conditioned on text and\/or other images as input through the diffusion process. Please refer to https:\/\/arxiv.org\/abs\/2112.10752 for details.",
+    "metadataOutputVersion" : "3.0",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float32",
+        "formattedType" : "MultiArray (Float32 1 × 8 × 64 × 64)",
+        "shortDescription" : "The latent embeddings from the unet model from the input image.",
+        "shape" : "[1, 8, 64, 64]",
+        "name" : "latent",
+        "type" : "MultiArray"
+      }
+    ],
+    "version" : "clover-image-tiny-inpaint-v2",
+    "modelParameters" : [
+
+    ],
+    "author" : "Please refer to the Model Card available at huggingface.co\/clover-image-tiny-inpaint-v2",
+    "specificationVersion" : 9,
+    "storagePrecision" : "Float16",
+    "license" : "OpenRAIL (https:\/\/huggingface.co\/spaces\/CompVis\/stable-diffusion-license)",
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.square" : 22,
+      "Ios18.batchNorm" : 21,
+      "Ios18.mul" : 1,
+      "Ios18.sub" : 22,
+      "Ios18.conv" : 28,
+      "Ios18.linear" : 4,
+      "Ios18.scaledDotProductAttention" : 1,
+      "Ios18.add" : 34,
+      "Ios18.realDiv" : 22,
+      "Ios18.silu" : 21,
+      "Ios16.reduceMean" : 44,
+      "Ios18.transpose" : 6,
+      "Ios18.cast" : 1,
+      "Ios18.reshape" : 49,
+      "Ios18.sqrt" : 22,
+      "Pad" : 3
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "stateSchema" : [
+
+    ],
+    "isUpdatable" : "0",
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 3 × 512 × 512)",
+        "shortDescription" : "The input image to base the initial latents on normalized to range [-1, 1]",
+        "shape" : "[1, 3, 512, 512]",
+        "name" : "x",
+        "type" : "MultiArray"
+      }
+    ],
+    "userDefinedMetadata" : {
+      "com.github.apple.coremltools.conversion_date" : "2026-08-11",
+      "com.github.apple.coremltools.source" : "torch==2.7.0",
+      "com.github.apple.coremltools.version" : "9.0",
+      "com.github.apple.coremltools.source_dialect" : "TorchScript"
+    },
+    "generatedClassName" : "Stable_Diffusion_version_clover_image_tiny_inpaint_v2_vae_encoder",
+    "method" : "predict"
+  }
+]

Resources/VAEEncoder.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71bf826e5ee1b455462aff0b34e3a469f5f58955ae715708962b1d81db92dc64
+size 68338112

manifest.json

@@ -21,21 +21,28 @@
     ],
     "mask_semantics": "white=regenerate, black=preserve"
   },
+  "recommended_inference": {
+    "scheduler": "dpm_solver_multistep",
+    "steps": 20,
+    "guidance_scale": 6.0,
+    "mask_crop_padding": 96,
+    "composite_outside_mask": "exact source pixels"
+  },
   "resources": [
     {
       "path": "SafetyChecker.mlmodelc/analytics/coremldata.bin",
       "size": 243,
-      "sha256": "2ce84e71c14d26c8d655425ebd3299692f1948e213228e9f427f9d7831cc5963"
+      "sha256": "8211508a22331e6868fa9d207c6532ae5fc11791e8ceac6bf6fa4c7dd89004e8"
     },
     {
       "path": "SafetyChecker.mlmodelc/coremldata.bin",
-      "size": 1543,
-      "sha256": "7425b537a536549ab5db59f87b7cde670cfaf1636a64f424710cbeeb70b893b0"
+      "size": 1549,
+      "sha256": "f57494e9bc01a089c73ba6b1b9554e1a3df155efbf7cf36223ba10ae494f0d58"
     },
     {
       "path": "SafetyChecker.mlmodelc/metadata.json",
-      "size": 4678,
-      "sha256": "2c33a16a32bcbb3ab175a6af1ff6bfff47a49ea91944e9d1eac15268f0ea6b8a"
+      "size": 4687,
+      "sha256": "4af6e1bd32288fd5e6e6f171d7e63a8a3f3ec776b45c16bb7e3fe352e36158f0"
     },
     {
       "path": "SafetyChecker.mlmodelc/model.mil",
@@ -50,92 +57,92 @@
     {
       "path": "TextEncoder.mlmodelc/analytics/coremldata.bin",
       "size": 243,
-      "sha256": "933b2c64d95e0816d52c0b77bd714a0f6e6a2ce7f11219521d8f130ab6adc696"
+      "sha256": "3ba6d80c5e94f50f17a70a965c2f63d24fa4239672d0fb5dadcf888e69cafc0f"
     },
     {
       "path": "TextEncoder.mlmodelc/coremldata.bin",
-      "size": 946,
-      "sha256": "4ce87b6dd6d848c36fefaf1a6808a1bf80839f3eb8d621eb8fdcb6a1077bc5f4"
+      "size": 952,
+      "sha256": "734dc3fdcdcc31367ffae6a58b1e4b0d713eb3f043b89811b7fcaec6df67a495"
     },
     {
       "path": "TextEncoder.mlmodelc/metadata.json",
-      "size": 2983,
-      "sha256": "e2a9ea1cfb6ee57e0686cfc9c1fd153cb39edb27c5731c9225c31fff9e58103b"
+      "size": 2992,
+      "sha256": "108a55b2e93052a093256052dbdfa9a6515c8918fbcd0c75661ffd753401ca25"
     },
     {
       "path": "TextEncoder.mlmodelc/model.mil",
-      "size": 152360,
-      "sha256": "6db8e2498aa8a1654f19d2ccc74643e1a503c95b04a1c1dd048d5da01d01fd72"
+      "size": 154795,
+      "sha256": "e3a95541282de8bc236c672d8170bfea3aceb478bd5fc5f96f390857d758be80"
     },
     {
       "path": "TextEncoder.mlmodelc/weights/weight.bin",
-      "size": 244957824,
-      "sha256": "79b9260c79cbb5b3ed22088fd2765aa5fc1b4b1c24a688537c2439276798ab0c"
+      "size": 246145536,
+      "sha256": "b3e8699e18a386ba82633384e0a6f5c090983f1c66b24a26fad3554fc2a12ddb"
     },
     {
       "path": "UnetChunk1.mlmodelc/analytics/coremldata.bin",
       "size": 243,
-      "sha256": "1c79e4e6107a0878b51625390eb1389466e2df8b0ca2436f477dc6a2935484c9"
+      "sha256": "7ee3483e4486c1bdc21299013aa9d3a25fcafb3d064750294c822cf8d6708151"
     },
     {
       "path": "UnetChunk1.mlmodelc/coremldata.bin",
       "size": 532,
-      "sha256": "4f78341f190d8fd01269c31f7d0a7576f3df9cafa007f7dcf47915d9d598d4da"
+      "sha256": "f1ebbe33a5324d71a517efe9d5f5b8f9bbcb9bd2e991a57171840161b3a63430"
     },
     {
       "path": "UnetChunk1.mlmodelc/metadata.json",
-      "size": 5239,
-      "sha256": "f22c332b1dbb9291668e1c34b233d9f5738acac3cef88d40fc450d3bf745c00f"
+      "size": 5242,
+      "sha256": "54c3b23ff2d46f00cbe5e3b03e79abf4e3fbb2bc811fd440219107c79dde2ef1"
     },
     {
       "path": "UnetChunk1.mlmodelc/model.mil",
       "size": 359479,
-      "sha256": "3cb66d8d6df9d9476b97f4c7058f80a20d74cc3d6a7930e47eb4d525b18babf1"
+      "sha256": "6067c2cbbb7f35ca3f3811ded8191fab58ddd1c27ee3de5e8249fa6e9c0d7581"
     },
     {
       "path": "UnetChunk1.mlmodelc/weights/weight.bin",
       "size": 324651584,
-      "sha256": "15d72e32d29f8aa5813a0bc9fca0ccdf1063411b91c496ec20fa6a25af8fdc13"
+      "sha256": "091529bc9e2b2bdda0508614e223570f7a8a4fe288bc510ec844635ba2f01046"
     },
     {
       "path": "UnetChunk2.mlmodelc/analytics/coremldata.bin",
       "size": 243,
-      "sha256": "3a468503aee82948546a734364424194ba66fe01ca5d4fd403fc75f7569234ed"
+      "sha256": "1f7d7cba85c9edd629d133c0224baf71e989ad4bfe1da89f37706f1d6cab45ea"
     },
     {
       "path": "UnetChunk2.mlmodelc/coremldata.bin",
       "size": 513,
-      "sha256": "45154abdc0829837df1817882dc4d6fe759241b60a55a316f735785999e3daac"
+      "sha256": "94ea7e153dab4f55ad8a40d12f9853422f8bdce240f5b608f3ae72ac3229fb35"
     },
     {
       "path": "UnetChunk2.mlmodelc/metadata.json",
-      "size": 4923,
-      "sha256": "07c071a911c066038501b098ee4fb0aef6cd1f3ccd9f0bb81053205825252bff"
+      "size": 4926,
+      "sha256": "4c3b0ad1527a1b457d22a70385ad3d8424d72d4cecb825a12e996539de649a16"
     },
     {
       "path": "UnetChunk2.mlmodelc/model.mil",
-      "size": 460762,
-      "sha256": "4461145daa177da05102d3dfba042486a329243dce7681644e1561d9fe243410"
+      "size": 460764,
+      "sha256": "82541a2df3d6b955984b151f66dc7b48f44773c1c4a3ad81b64c474bea56ba9e"
     },
     {
       "path": "UnetChunk2.mlmodelc/weights/weight.bin",
       "size": 322211328,
-      "sha256": "06720754c9b2d5f1165a7989f07be3cb8a879f68363b27f85820afb47ce9353d"
+      "sha256": "04910ba6d4ca88b34dd1202c848aebfc33e339010d203573ebc4138a4e9f78ba"
     },
     {
       "path": "VAEDecoder.mlmodelc/analytics/coremldata.bin",
       "size": 243,
-      "sha256": "dd5e8ee0e1e8766ed20fc56a213be2a7319998d4603fad7d5d3fbf11f74b0fea"
+      "sha256": "ff20b000564aa02c66b7e254258a5c8ba7d34fdb99c54be3c85a42d8eea45a62"
     },
     {
       "path": "VAEDecoder.mlmodelc/coremldata.bin",
-      "size": 873,
-      "sha256": "b5871a4489e3fe7daf6244257e715f27c55281affea5d6fa0de0caefc8092615"
+      "size": 879,
+      "sha256": "7bb1f9ddd844d6c8e14c6e52a1e05d7c493e71091d788798d065141bbe3e131b"
     },
     {
       "path": "VAEDecoder.mlmodelc/metadata.json",
-      "size": 2681,
-      "sha256": "cd013988d5b68a0f4046b0ba8faee68b0a73b8bf9749a7d18497907feffd6b19"
+      "size": 2690,
+      "sha256": "72a20fd0a1db9e6d3395c6be9ab3c91b70fb771c138d8b7fef5ccc6b9153539c"
     },
     {
       "path": "VAEDecoder.mlmodelc/model.mil",
@@ -150,17 +157,17 @@
     {
       "path": "VAEEncoder.mlmodelc/analytics/coremldata.bin",
       "size": 243,
-      "sha256": "e060c88a7c335fc597b5a583d7e43da35ab398819dbfed9e57f5a02c28cffb15"
+      "sha256": "ba122c04ae4c8d28c8b8805049e22fb7dfeac3eada7bde68c2366a14e712d1d6"
     },
     {
       "path": "VAEEncoder.mlmodelc/coremldata.bin",
-      "size": 877,
-      "sha256": "e4a4dfa8b3370da5e69a7ad0c14d06b3e1ce550a32ccc8dabb1f60e68ae53d0c"
+      "size": 883,
+      "sha256": "fe1d98a943015c7ac3d29ef2d573b37610e62c97e5724d2f85c555960832629d"
     },
     {
       "path": "VAEEncoder.mlmodelc/metadata.json",
-      "size": 2665,
-      "sha256": "6e18503b05891fc290b3b25d1c42afac362828199695eeee2c1abb6c4ba485df"
+      "size": 2674,
+      "sha256": "f18ce0aabcf1ad6695c8d7bf418dd30bdac1d031ef453fe1f2b655487dff019d"
     },
     {
       "path": "VAEEncoder.mlmodelc/model.mil",
@@ -179,8 +186,8 @@
     },
     {
       "path": "vocab.json",
-      "size": 1059962,
-      "sha256": "e089ad92ba36837a0d31433e555c8f45fe601ab5c221d4f607ded32d9f7a4349"
+      "size": 1098502,
+      "sha256": "4e6a66b6903c12d273a1a6645f761a2de07e239facfc52fd797b83a75014ca78"
     }
   ]
 }