Add files using upload-large-folder tool

CoreMLModels/DecoderChunk0Decode.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:43bae48dccc93879c6c6faa1de5706e2de19127a846f67ad72594099610ff421
+oid sha256:0cc387026ed735ea609b615601027aa83792d0febaef4dd6f0e349628b853c16
 size 64555

CoreMLModels/DecoderChunk0Decode.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "6AA1FB5C-E899-474E-A862-007153C8CAA4": {
+        "1A98F518-88C5-4C92-8883-CBAFCDF2320D": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         },
-        "ED22A1A7-60AC-4E18-B171-8F3FA6C8BD9C": {
+        "DF5FBF20-A19D-46E2-A2E3-C5712D14EF9F": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "6AA1FB5C-E899-474E-A862-007153C8CAA4"
+    "rootModelIdentifier": "1A98F518-88C5-4C92-8883-CBAFCDF2320D"
 }

CoreMLModels/DecoderChunk0Prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ebdc940d1c2e987c1ca92becece863a2f8b503ee87d8a751a1baca1b591ffe14
+oid sha256:33a9637acec9d3d604e199b7586c0802577af10c31a31b44a0b3d9be47f3bdda
 size 64778

CoreMLModels/DecoderChunk0Prefill.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "2D215524-918F-4769-8994-6B2FE905F2E0": {
+        "43E89C25-E73C-4E27-89B0-E1E2DED9FC0A": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "DCB6626C-17AE-450C-B6A5-708353132A5D": {
+        "A071D5EF-6D62-4775-868F-DB6862311E2C": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "DCB6626C-17AE-450C-B6A5-708353132A5D"
+    "rootModelIdentifier": "A071D5EF-6D62-4775-868F-DB6862311E2C"
 }

CoreMLModels/DecoderChunk1Decode.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6ee08d3f28600333ec167978635e53d524f04cca3780e025bead3c034b26790
+oid sha256:b229b0189a46e8f0e28e2b3f8febe0b214e8c5f525ab513b19b3813c811a6ff3
 size 64548

CoreMLModels/DecoderChunk1Decode.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "94DB4639-740C-45A4-B5BA-0ADD4A615074": {
+        "DFFED78D-B2E8-4469-8472-A6A2CCF6FFC7": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         },
-        "FEBDF55D-E589-470A-B617-B496CF2A9C53": {
+        "E90FCD54-BEBA-46DC-B5B4-5E2A193E6073": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "94DB4639-740C-45A4-B5BA-0ADD4A615074"
+    "rootModelIdentifier": "DFFED78D-B2E8-4469-8472-A6A2CCF6FFC7"
 }

CoreMLModels/DecoderChunk1Prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b112d1c3862610d4b822de60187771d563ba85085063289265c1aa0fb206e5bf
+oid sha256:520d7df9e15bf41ecec0dd39f85028f0bcfaf3d06491db472a9f675931079323
 size 64770

CoreMLModels/DecoderChunk1Prefill.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "9E9438C8-3C51-49E4-A7F4-46C0AC6712E7": {
+        "470A72DD-A10E-4679-AEE1-C4C85C814E51": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         },
-        "E3F7C9C8-4262-4A72-9156-3B60F6593929": {
+        "8ECB37DB-0876-4CF2-9D94-814800D1056C": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "9E9438C8-3C51-49E4-A7F4-46C0AC6712E7"
+    "rootModelIdentifier": "470A72DD-A10E-4679-AEE1-C4C85C814E51"
 }

CoreMLModels/DecoderChunk2Decode.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30281ca628f916fa54bd274e51a535abfc39ba8b98049b0e854dd57dbe2d0958
+oid sha256:d3b2ebc592e6f854e0abee328363d94dcf1a61fddf410309a25b0a9dc971d991
 size 79108

CoreMLModels/DecoderChunk2Decode.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "486149AD-860C-48C5-8E77-BFA452929FC2": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Specification",
-            "name": "model.mlmodel",
-            "path": "com.apple.CoreML/model.mlmodel"
-        },
-        "88F37AD6-23D7-4467-A029-7C354C6C8884": {
+        "41A3EA4C-2696-44B5-9D35-F63D4C09D970": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
+        },
+        "C73F7874-F814-4C95-AE95-02ED9F074D7A": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "486149AD-860C-48C5-8E77-BFA452929FC2"
+    "rootModelIdentifier": "C73F7874-F814-4C95-AE95-02ED9F074D7A"
 }

CoreMLModels/DecoderChunk2Prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d9dc23c97baa9e535e9042eaf363567842e4ff996a962cfad3e07a93dcf1d988
+oid sha256:dc07cf8a587117eeacb0ab740eaab773d00971bbc4620c274816196e25b46d94
 size 79361

CoreMLModels/DecoderChunk2Prefill.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "0CB73A89-FACA-4C40-ABB2-B1AC212D3AE9": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Specification",
-            "name": "model.mlmodel",
-            "path": "com.apple.CoreML/model.mlmodel"
-        },
-        "91D4E742-98AC-4579-979B-A288B7B07B9D": {
+        "0134A2B6-3755-4199-AAC2-66D1AC412C4E": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
+        },
+        "CD529845-4038-4403-8061-A18DC34452F2": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "0CB73A89-FACA-4C40-ABB2-B1AC212D3AE9"
+    "rootModelIdentifier": "CD529845-4038-4403-8061-A18DC34452F2"
 }

CoreMLModels/DecoderChunk3Decode.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c8909d20993025453e1327dc778f543526b7b60680bba03924458924150558b
+oid sha256:af3395fbb7c39e29f787d5ee2dffcf8b95769793a4fdf15b924a94a938727f35
 size 80877

CoreMLModels/DecoderChunk3Decode.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "1E34A255-F2AA-4F18-A25B-31F908E88716": {
+        "4AE037AC-8752-4645-95B4-41B590B7B38A": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "DD824B66-DEFB-49A8-8B07-C95628029584": {
+        "AEFA37C9-987A-4D88-B436-1B1D0DAA5EC7": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "DD824B66-DEFB-49A8-8B07-C95628029584"
+    "rootModelIdentifier": "AEFA37C9-987A-4D88-B436-1B1D0DAA5EC7"
 }

CoreMLModels/DecoderChunk3Prefill.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89d1e48be73e55d143395541a3f7045e0a53ec9a5f2012cf3a9da9425ba6a534
+oid sha256:07c4ae3638c84b5283ece69f2a8a19b03d6ab28ec5d3ff120d4769a180800c80
 size 81136

CoreMLModels/DecoderChunk3Prefill.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "2B444598-A85E-406C-82E4-0A6B2F13E057": {
+        "0F94A726-F6CC-4E4E-A5E8-AC048AFCE89F": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "BF42EB5C-FA49-44CB-8746-DAC7C10F4974": {
+        "BD25D75A-69D4-43B4-9920-9A475D358178": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "BF42EB5C-FA49-44CB-8746-DAC7C10F4974"
+    "rootModelIdentifier": "BD25D75A-69D4-43B4-9920-9A475D358178"
 }

CoreMLModels/Embeddings.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3161f212eede005d03a7423002b281d34cd8d77fbfcfa3b7cbf8947e94097371
+oid sha256:816273dca321b72310418d1af0408316b5cafdd812c0a8c130b12cec68605ced
 size 1939

CoreMLModels/Embeddings.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "45BE8FB8-0AB1-4544-9C62-AE155F9CCC07": {
-            "author": "com.apple.CoreML",
-            "description": "CoreML Model Weights",
-            "name": "weights",
-            "path": "com.apple.CoreML/weights"
-        },
-        "D055084C-4E6A-4522-B6C9-9A7D732884BA": {
+        "6A233700-5976-490E-992F-8978689138D8": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "883BBA59-0DEA-4550-A4EE-B5E812264CB0": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
         }
     },
-    "rootModelIdentifier": "D055084C-4E6A-4522-B6C9-9A7D732884BA"
+    "rootModelIdentifier": "6A233700-5976-490E-992F-8978689138D8"
 }

CoreMLModels/LMHead.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab815fd22169d59bf66d5bdfffa82463d4baf4200ddab419e034378d00fd0389
+oid sha256:8f125cfa030aa1d65d27adebf2afb151129d808002a0b902926175602e592b05
 size 1181

CoreMLModels/LMHead.mlpackage/Manifest.json

@@ -1,18 +1,18 @@
 {
     "fileFormatVersion": "1.0.0",
     "itemInfoEntries": {
-        "0A0BFB9F-E052-41FE-97EF-863F42591C7E": {
+        "5A0544CF-EBEF-4899-91CA-7CC14764EA18": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Weights",
             "name": "weights",
             "path": "com.apple.CoreML/weights"
         },
-        "B0B497AC-AAD8-4945-B42F-82455C3A3A54": {
+        "AD9FD3DA-D884-4155-B089-D02E6DC81573": {
             "author": "com.apple.CoreML",
             "description": "CoreML Model Specification",
             "name": "model.mlmodel",
             "path": "com.apple.CoreML/model.mlmodel"
         }
     },
-    "rootModelIdentifier": "B0B497AC-AAD8-4945-B42F-82455C3A3A54"
+    "rootModelIdentifier": "AD9FD3DA-D884-4155-B089-D02E6DC81573"
 }

CoreMLModels/VisionEncoder.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15518121cbaadac79d610d0f9d7791455310e817f2943afbab350aa45784004e
+size 280141

CoreMLModels/VisionEncoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93a3c45689837caec7d759884c00a58d3b3d3acc2b3628f5359e410f99fdff27
+size 822966528

CoreMLModels/VisionEncoder.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "C70439C3-5557-4551-8388-7A0549257A0E": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "F8C433C7-A367-45BC-9F28-D8A1B03850BD": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        }
+    },
+    "rootModelIdentifier": "C70439C3-5557-4551-8388-7A0549257A0E"
+}

CoreMLModels/VisionPatchEmbedding.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:265cc9efda8502e433ec6354e34a1efb5ad35ae0e499382e145b253dc897f398
+size 1158

CoreMLModels/VisionPatchEmbedding.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e06bbf3458c4288524f73043ff0add87494039d297b2d302b602d0d27a25d25e
+size 1771968

CoreMLModels/VisionPatchEmbedding.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "347CC389-114C-43F8-8B63-F9C3AB451B47": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "F8B7B706-9212-40E6-8D25-1DBD536BD86A": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        }
+    },
+    "rootModelIdentifier": "347CC389-114C-43F8-8B63-F9C3AB451B47"
+}

CoreMLModels/VisionPositionEmbeddings.float16.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7aba5ca4d8eb987174368bf00374ea5484fa30aabc14bba44a78282d7f0ee21f
+size 589824

CoreMLModels/VisionProjector.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc1f116ad04e879e9d0d0e4e82bb9bb550710fa069e4ed360753c8f2355150bf
+size 6830

CoreMLModels/VisionProjector.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e848d403fadcbd3448095a56483fa2c239700effb5f61121350c4464f33ce98
+size 27271488

CoreMLModels/VisionProjector.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "12FD411A-85DB-4CA9-8DD1-78E37AA0988A": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "65A8478D-D9CC-42BC-B0F8-15FE36FA149E": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        }
+    },
+    "rootModelIdentifier": "12FD411A-85DB-4CA9-8DD1-78E37AA0988A"
+}

CoreMLModels/meta.json

@@ -14,6 +14,23 @@
   "image_token_id": 396,
   "image_start_token_id": 498,
   "image_end_token_id": 499,
+  "image_thumbnail_token_id": 497,
+  "tile_size": 512,
+  "encoder_patch_size": 16,
+  "downsample_factor": 2,
+  "min_tiles": 2,
+  "max_tiles": 10,
+  "use_thumbnail": true,
+  "min_image_tokens": 64,
+  "max_image_tokens": 256,
+  "max_num_patches": 1024,
+  "vision_hidden_size": 1152,
+  "projector_hidden_size": 2048,
+  "vision_position_embedding_size": 16,
+  "vision_patch_embedding_model": "VisionPatchEmbedding.mlpackage",
+  "vision_encoder_model": "VisionEncoder.mlpackage",
+  "vision_projector_model": "VisionProjector.mlpackage",
+  "vision_position_embeddings_file": "VisionPositionEmbeddings.float16.bin",
   "chunk_order": [
     "DecoderChunk0",
     "DecoderChunk1",

README.md

@@ -9,14 +9,16 @@ tags:
 - ane
 - swift
 - text-generation
+- multimodal
+- vision-language
 - liquid
 - lfm2.5-vl
-pipeline_tag: text-generation
+pipeline_tag: image-text-to-text
 ---
 
 # LFM2.5-VL-1.6B CoreML (ANE Native)
 
-CoreML ML Program export of the LFM2.5-VL-1.6B text decoder, packaged for Apple Silicon with a fixed 4096-token context window.
+CoreML ML Program export of LFM2.5-VL-1.6B for Apple Silicon with a fixed 4096-token context window, including the text decoder plus split multimodal vision assets.
 
 Swift inference package and CLI: https://github.com/mweinbach/LFM2.5-VL-1.6B-ANE-Native
 
@@ -26,6 +28,10 @@ Swift inference package and CLI: https://github.com/mweinbach/LFM2.5-VL-1.6B-ANE
 - `CoreMLModels/LMHead.mlpackage`
 - `CoreMLModels/DecoderChunk{0-3}Prefill.mlpackage`
 - `CoreMLModels/DecoderChunk{0-3}Decode.mlpackage`
+- `CoreMLModels/VisionPatchEmbedding.mlpackage`
+- `CoreMLModels/VisionEncoder.mlpackage`
+- `CoreMLModels/VisionProjector.mlpackage`
+- `CoreMLModels/VisionPositionEmbeddings.float16.bin`
 - `CoreMLModels/meta.json`
 - `tokenizer.json`
 - `tokenizer_config.json`
@@ -38,18 +44,16 @@ Swift inference package and CLI: https://github.com/mweinbach/LFM2.5-VL-1.6B-ANE
 
 - 4096-token fixed context window
 - chunked decoder export for prompt prefill + single-token decode
+- split vision path: patch embedding + encoder + projector
+- supports multi-image prompts and the model's tiling/thumbnail preprocessing flow
 - Apple CoreML / ML Program target
 - tokenizer and chat template included for local execution
 
-## Current scope
-
-- text-generation runtime only
-- no vision encoder/projector export in this package yet
-- intended to be used with the matching Swift package linked above
-
 ## Usage
 
-Clone or download this repo, then point the Swift CLI at the downloaded folder:
+Clone or download this repo, then point the Swift CLI at the downloaded folder.
+
+Text-only example:
 
 ```bash
 swift run ANEInferenceCLI \
@@ -58,5 +62,15 @@ swift run ANEInferenceCLI \
   --max-new-tokens 64
 ```
 
-The Swift package is here:
+Multimodal example:
+
+```bash
+swift run ANEInferenceCLI \
+  --bundle-root /path/to/LFM2.5-VL-1.6B-CoreML \
+  --image /path/to/image.png \
+  --prompt "Describe the image briefly." \
+  --max-new-tokens 64
+```
+
+The matching Swift package is here:
 https://github.com/mweinbach/LFM2.5-VL-1.6B-ANE-Native