Initial upload: 3-chunk merged stateful Linear (Stage 3)

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+hf_model/tokenizer.json filter=lfs diff=lfs merge=lfs -text

chunk_1.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d96a0f6b2d6273ed0cfc277a52481c869b7f169b1df15b37c8974bb7f086976e
+size 243

chunk_1.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07bc110005a3c9437dafced7453d31c8e1ad53ddc95024de7f4362ede80c132a
+size 1009

chunk_1.mlmodelc/metadata.json

@@ -0,0 +1,554 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (12 bits), Palettized (13 bits), Palettized (7 bits), Palettized (8 bits), UInt4)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states_out",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 8960]",
+        "name" : "per_layer_combined_out",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask_full",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "causal_mask_sliding",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8960]",
+            "name" : "per_layer_raw",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 256]",
+            "name" : "cos_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 256]",
+            "name" : "sin_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "cos_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "sin_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "ring_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Float32, Int32)",
+        "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (12 bits), Palettized (13 bits), Palettized (7 bits), Palettized (8 bits), UInt4)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 14 × 1 × 512 × 512)",
+            "shortDescription" : "",
+            "shape" : "[14, 1, 512, 512]",
+            "name" : "kv_cache_sliding",
+            "type" : "State"
+          },
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 2 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[2, 1, 2048, 512]",
+            "name" : "kv_cache_full",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states_out",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8960]",
+            "name" : "per_layer_combined_out",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.mul" : 196,
+          "Ios18.matmul" : 16,
+          "Ios18.rsqrt" : 8,
+          "Ios16.reduceMean" : 8,
+          "Ios18.exp" : 8,
+          "Ios18.realDiv" : 8,
+          "Split" : 73,
+          "Ios18.readState" : 18,
+          "Ios16.reduceMax" : 8,
+          "Ios18.add" : 59,
+          "Tile" : 16,
+          "Ios18.layerNorm" : 57,
+          "Ios18.sliceUpdate" : 16,
+          "Ios18.writeState" : 16,
+          "Pad" : 14,
+          "Ios18.reshape" : 98,
+          "Ios16.reduceSum" : 8,
+          "Ios18.constexprLutToDense" : 73,
+          "Ios18.linear" : 73,
+          "Ios18.concat" : 105,
+          "Ios18.transpose" : 80,
+          "Ios18.sub" : 8,
+          "Ios18.pow" : 8,
+          "Ios18.gelu" : 16,
+          "Ios18.sliceByIndex" : 38
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 2048]",
+            "name" : "causal_mask_full",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "causal_mask_sliding",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 8960]",
+            "name" : "per_layer_raw",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 256]",
+            "name" : "cos_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 256]",
+            "name" : "sin_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "cos_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "sin_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "ring_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Float32, Int32)",
+        "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (12 bits), Palettized (13 bits), Palettized (7 bits), Palettized (8 bits), UInt4)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 14 × 1 × 512 × 512)",
+            "shortDescription" : "",
+            "shape" : "[14, 1, 512, 512]",
+            "name" : "kv_cache_sliding",
+            "type" : "State"
+          },
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 2 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[2, 1, 2048, 512]",
+            "name" : "kv_cache_full",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 1536]",
+            "name" : "hidden_states_out",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 8960]",
+            "name" : "per_layer_combined_out",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill_b8",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.mul" : 196,
+          "Ios18.matmul" : 16,
+          "Ios18.rsqrt" : 8,
+          "Ios16.reduceMean" : 8,
+          "Ios18.exp" : 8,
+          "Ios18.realDiv" : 8,
+          "Split" : 73,
+          "Ios18.readState" : 18,
+          "Ios16.reduceMax" : 8,
+          "Ios18.add" : 59,
+          "Tile" : 16,
+          "Ios18.layerNorm" : 57,
+          "Ios18.sliceUpdate" : 16,
+          "Ios18.writeState" : 16,
+          "Pad" : 14,
+          "Ios18.reshape" : 66,
+          "Ios16.reduceSum" : 8,
+          "Ios18.constexprLutToDense" : 73,
+          "Ios18.linear" : 73,
+          "Ios18.concat" : 105,
+          "Ios18.transpose" : 80,
+          "Ios18.sub" : 8,
+          "Ios18.pow" : 8,
+          "Ios18.gelu" : 16,
+          "Ios18.sliceByIndex" : 38
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.mul" : 196,
+      "Ios18.matmul" : 16,
+      "Ios18.rsqrt" : 8,
+      "Ios16.reduceMean" : 8,
+      "Ios18.exp" : 8,
+      "Ios18.realDiv" : 8,
+      "Split" : 73,
+      "Ios18.readState" : 18,
+      "Ios16.reduceMax" : 8,
+      "Ios18.add" : 59,
+      "Tile" : 16,
+      "Ios18.layerNorm" : 57,
+      "Ios18.sliceUpdate" : 16,
+      "Ios18.writeState" : 16,
+      "Pad" : 14,
+      "Ios18.reshape" : 98,
+      "Ios16.reduceSum" : 8,
+      "Ios18.constexprLutToDense" : 73,
+      "Ios18.linear" : 73,
+      "Ios18.concat" : 105,
+      "Ios18.transpose" : 80,
+      "Ios18.sub" : 8,
+      "Ios18.pow" : 8,
+      "Ios18.gelu" : 16,
+      "Ios18.sliceByIndex" : 38
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 14 × 1 × 512 × 512)",
+        "shortDescription" : "",
+        "shape" : "[14, 1, 512, 512]",
+        "name" : "kv_cache_sliding",
+        "type" : "State"
+      },
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 2 × 1 × 2048 × 512)",
+        "shortDescription" : "",
+        "shape" : "[2, 1, 2048, 512]",
+        "name" : "kv_cache_full",
+        "type" : "State"
+      }
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask_full",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "causal_mask_sliding",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 8960]",
+        "name" : "per_layer_raw",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 256]",
+        "name" : "cos_s",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 256]",
+        "name" : "sin_s",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "cos_f",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "sin_f",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "current_pos",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "ring_pos",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "chunk_1",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

chunk_1.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dbebbe38ca34dad177f3e9123e1dfb89b0d5b6fac49b3b52868f7c50b08f0cd
+size 155484416

chunk_2.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34163deded8f46eb55c7c05b11036257c91d5262ae348dcc8fb699869261e674
+size 243

chunk_2.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80547a545eb58019882825674410c3084221805dc0b4c5f275deb3e32325e0ed
+size 1147

chunk_2.mlmodelc/metadata.json

@@ -0,0 +1,644 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (12 bits), Palettized (13 bits), Palettized (7 bits), Palettized (8 bits), UInt4)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states_out",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 512, 256]",
+        "name" : "kv13_k",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 512, 256]",
+        "name" : "kv13_v",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 2048, 512]",
+        "name" : "kv14_k",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 2048, 512]",
+        "name" : "kv14_v",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask_full",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "causal_mask_sliding",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8960]",
+            "name" : "per_layer_combined",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 256]",
+            "name" : "cos_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 256]",
+            "name" : "sin_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "cos_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "sin_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "ring_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Float32, Int32)",
+        "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (12 bits), Palettized (13 bits), Palettized (7 bits), Palettized (8 bits), UInt4)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 10 × 1 × 512 × 512)",
+            "shortDescription" : "",
+            "shape" : "[10, 1, 512, 512]",
+            "name" : "kv_cache_sliding",
+            "type" : "State"
+          },
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 4 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[4, 1, 2048, 512]",
+            "name" : "kv_cache_full",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states_out",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_v",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_v",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.mul" : 347,
+          "Ios18.matmul" : 34,
+          "Ios18.rsqrt" : 7,
+          "Ios16.reduceMean" : 7,
+          "Ios18.exp" : 17,
+          "Ios18.realDiv" : 17,
+          "Split" : 133,
+          "Ios18.readState" : 16,
+          "Ios16.reduceMax" : 17,
+          "Ios18.add" : 101,
+          "Tile" : 14,
+          "Ios18.layerNorm" : 109,
+          "Ios18.sliceUpdate" : 14,
+          "Ios18.writeState" : 14,
+          "Pad" : 10,
+          "Ios18.reshape" : 124,
+          "Ios16.reduceSum" : 17,
+          "Ios18.constexprLutToDense" : 133,
+          "Ios18.linear" : 133,
+          "Ios18.concat" : 161,
+          "Ios18.transpose" : 90,
+          "Ios18.sub" : 17,
+          "Ios18.pow" : 7,
+          "Ios18.gelu" : 34,
+          "Ios18.sliceByIndex" : 41
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 2048]",
+            "name" : "causal_mask_full",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 �� 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "causal_mask_sliding",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 8960]",
+            "name" : "per_layer_combined",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 256]",
+            "name" : "cos_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 256]",
+            "name" : "sin_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "cos_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "sin_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "current_pos",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "ring_pos",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Float32, Int32)",
+        "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (12 bits), Palettized (13 bits), Palettized (7 bits), Palettized (8 bits), UInt4)",
+        "stateSchema" : [
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 10 × 1 × 512 × 512)",
+            "shortDescription" : "",
+            "shape" : "[10, 1, 512, 512]",
+            "name" : "kv_cache_sliding",
+            "type" : "State"
+          },
+          {
+            "dataType" : "Float16",
+            "isOptional" : "0",
+            "formattedType" : "State (Float16 4 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[4, 1, 2048, 512]",
+            "name" : "kv_cache_full",
+            "type" : "State"
+          }
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 1536]",
+            "name" : "hidden_states_out",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_v",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_v",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill_b8",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.mul" : 347,
+          "Ios18.matmul" : 34,
+          "Ios18.rsqrt" : 7,
+          "Ios16.reduceMean" : 7,
+          "Ios18.exp" : 17,
+          "Ios18.realDiv" : 17,
+          "Split" : 133,
+          "Ios18.readState" : 16,
+          "Ios16.reduceMax" : 17,
+          "Ios18.add" : 101,
+          "Tile" : 14,
+          "Ios18.layerNorm" : 109,
+          "Ios18.sliceUpdate" : 14,
+          "Ios18.writeState" : 14,
+          "Pad" : 10,
+          "Ios18.reshape" : 76,
+          "Ios16.reduceSum" : 17,
+          "Ios18.constexprLutToDense" : 133,
+          "Ios18.linear" : 133,
+          "Ios18.concat" : 161,
+          "Ios18.transpose" : 90,
+          "Ios18.sub" : 17,
+          "Ios18.pow" : 7,
+          "Ios18.gelu" : 34,
+          "Ios18.sliceByIndex" : 41
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.mul" : 347,
+      "Ios18.matmul" : 34,
+      "Ios18.rsqrt" : 7,
+      "Ios16.reduceMean" : 7,
+      "Ios18.exp" : 17,
+      "Ios18.realDiv" : 17,
+      "Split" : 133,
+      "Ios18.readState" : 16,
+      "Ios16.reduceMax" : 17,
+      "Ios18.add" : 101,
+      "Tile" : 14,
+      "Ios18.layerNorm" : 109,
+      "Ios18.sliceUpdate" : 14,
+      "Ios18.writeState" : 14,
+      "Pad" : 10,
+      "Ios18.reshape" : 124,
+      "Ios16.reduceSum" : 17,
+      "Ios18.constexprLutToDense" : 133,
+      "Ios18.linear" : 133,
+      "Ios18.concat" : 161,
+      "Ios18.transpose" : 90,
+      "Ios18.sub" : 17,
+      "Ios18.pow" : 7,
+      "Ios18.gelu" : 34,
+      "Ios18.sliceByIndex" : 41
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 10 × 1 × 512 × 512)",
+        "shortDescription" : "",
+        "shape" : "[10, 1, 512, 512]",
+        "name" : "kv_cache_sliding",
+        "type" : "State"
+      },
+      {
+        "dataType" : "Float16",
+        "isOptional" : "0",
+        "formattedType" : "State (Float16 4 × 1 × 2048 × 512)",
+        "shortDescription" : "",
+        "shape" : "[4, 1, 2048, 512]",
+        "name" : "kv_cache_full",
+        "type" : "State"
+      }
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Float32, Int32)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask_full",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "causal_mask_sliding",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 8960]",
+        "name" : "per_layer_combined",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 256]",
+        "name" : "cos_s",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 256]",
+        "name" : "sin_s",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "cos_f",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "sin_f",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "current_pos",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "ring_pos",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "chunk_2",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

chunk_2.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c14fb533fe553f98f35702aa320da35edafe1194d09a120635040fa3dc012200
+size 459299328

chunk_3.mlmodelc/analytics/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4754acc71a325a622165da06fb41f91208ec06b6cafb5ed196db46033a2e7f00
+size 243

chunk_3.mlmodelc/coremldata.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cec10f75acb7f2833c242c96a264165e7246cb166ff6ead0a4f6704a5011fba
+size 993

chunk_3.mlmodelc/metadata.json

@@ -0,0 +1,587 @@
+[
+  {
+    "metadataOutputVersion" : "3.0",
+    "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (13 bits), Palettized (17 bits), Palettized (7 bits), UInt4)",
+    "outputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Int32",
+        "formattedType" : "MultiArray (Int32 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "token_id",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1)",
+        "shortDescription" : "",
+        "shape" : "[1]",
+        "name" : "token_logit",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_normed",
+        "type" : "MultiArray"
+      }
+    ],
+    "modelParameters" : [
+
+    ],
+    "specificationVersion" : 9,
+    "functions" : [
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 2048]",
+            "name" : "causal_mask_full",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "causal_mask_sliding",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8960]",
+            "name" : "per_layer_combined",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 256]",
+            "name" : "cos_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 256]",
+            "name" : "sin_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "cos_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1, 512]",
+            "name" : "sin_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_v",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_v",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int32)",
+        "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (13 bits), Palettized (17 bits), Palettized (7 bits), UInt4)",
+        "stateSchema" : [
+
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "token_id",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "token_logit",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_normed",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "infer",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 1,
+          "Ios18.mul" : 184,
+          "Ios18.matmul" : 20,
+          "Ios18.exp" : 10,
+          "Ios18.realDiv" : 10,
+          "Split" : 71,
+          "Ios18.gatherAlongAxis" : 1,
+          "Ios16.reduceMax" : 10,
+          "Tile" : 4,
+          "Ios16.reduceSum" : 10,
+          "Ios18.add" : 50,
+          "Ios18.layerNorm" : 61,
+          "Ios18.reduceArgmax" : 1,
+          "Ios18.reshape" : 48,
+          "Ios18.constexprLutToDense" : 71,
+          "Ios18.linear" : 71,
+          "Ios18.gelu" : 20,
+          "Ios18.concat" : 71,
+          "Ios18.sub" : 10,
+          "Ios18.tanh" : 1,
+          "Ios18.transpose" : 32,
+          "Ios18.squeeze" : 2,
+          "Ios18.sliceByIndex" : 10
+        }
+      },
+      {
+        "inputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 1536]",
+            "name" : "hidden_states",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 2048)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 2048]",
+            "name" : "causal_mask_full",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "causal_mask_sliding",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 8 × 8960)",
+            "shortDescription" : "",
+            "shape" : "[1, 8, 8960]",
+            "name" : "per_layer_combined",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 256]",
+            "name" : "cos_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 256]",
+            "name" : "sin_s",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "cos_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 8 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 8, 512]",
+            "name" : "sin_f",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 512, 256]",
+            "name" : "kv13_v",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_k",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 2048, 512]",
+            "name" : "kv14_v",
+            "type" : "MultiArray"
+          }
+        ],
+        "computePrecision" : "Mixed (Float16, Int32)",
+        "storagePrecision" : "Mixed (Float16, Palettized (10 bits), Palettized (11 bits), Palettized (13 bits), Palettized (17 bits), Palettized (7 bits), UInt4)",
+        "stateSchema" : [
+
+        ],
+        "outputSchema" : [
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Int32",
+            "formattedType" : "MultiArray (Int32 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "token_id",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1)",
+            "shortDescription" : "",
+            "shape" : "[1]",
+            "name" : "token_logit",
+            "type" : "MultiArray"
+          },
+          {
+            "hasShapeFlexibility" : "0",
+            "isOptional" : "0",
+            "dataType" : "Float16",
+            "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+            "shortDescription" : "",
+            "shape" : "[1, 1, 1536]",
+            "name" : "hidden_normed",
+            "type" : "MultiArray"
+          }
+        ],
+        "name" : "prefill_b8",
+        "mlProgramOperationTypeHistogram" : {
+          "Ios18.expandDims" : 1,
+          "Ios18.mul" : 184,
+          "Ios18.matmul" : 20,
+          "Ios18.exp" : 10,
+          "Ios18.realDiv" : 10,
+          "Split" : 71,
+          "Ios18.gatherAlongAxis" : 1,
+          "Ios16.reduceMax" : 10,
+          "Tile" : 4,
+          "Ios16.reduceSum" : 10,
+          "Ios18.add" : 50,
+          "Ios18.layerNorm" : 61,
+          "Ios18.reduceArgmax" : 1,
+          "Ios18.reshape" : 28,
+          "Ios18.constexprLutToDense" : 71,
+          "Ios18.linear" : 71,
+          "Ios18.gelu" : 20,
+          "Ios18.concat" : 71,
+          "Ios18.sub" : 10,
+          "Ios18.tanh" : 1,
+          "Ios18.transpose" : 32,
+          "Ios18.squeeze" : 2,
+          "Ios18.sliceByIndex" : 11
+        }
+      }
+    ],
+    "mlProgramOperationTypeHistogram" : {
+      "Ios18.expandDims" : 1,
+      "Ios18.mul" : 184,
+      "Ios18.matmul" : 20,
+      "Ios18.exp" : 10,
+      "Ios18.realDiv" : 10,
+      "Split" : 71,
+      "Ios18.gatherAlongAxis" : 1,
+      "Ios16.reduceMax" : 10,
+      "Tile" : 4,
+      "Ios16.reduceSum" : 10,
+      "Ios18.add" : 50,
+      "Ios18.layerNorm" : 61,
+      "Ios18.reduceArgmax" : 1,
+      "Ios18.reshape" : 48,
+      "Ios18.constexprLutToDense" : 71,
+      "Ios18.linear" : 71,
+      "Ios18.gelu" : 20,
+      "Ios18.concat" : 71,
+      "Ios18.sub" : 10,
+      "Ios18.tanh" : 1,
+      "Ios18.transpose" : 32,
+      "Ios18.squeeze" : 2,
+      "Ios18.sliceByIndex" : 10
+    },
+    "isUpdatable" : "0",
+    "stateSchema" : [
+
+    ],
+    "availability" : {
+      "macOS" : "15.0",
+      "tvOS" : "18.0",
+      "visionOS" : "2.0",
+      "watchOS" : "11.0",
+      "iOS" : "18.0",
+      "macCatalyst" : "18.0"
+    },
+    "computePrecision" : "Mixed (Float16, Int32)",
+    "modelType" : {
+      "name" : "MLModelType_mlProgram"
+    },
+    "inputSchema" : [
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1536)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1536]",
+        "name" : "hidden_states",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 2048]",
+        "name" : "causal_mask_full",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "causal_mask_sliding",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 8960)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 8960]",
+        "name" : "per_layer_combined",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 256]",
+        "name" : "cos_s",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 256]",
+        "name" : "sin_s",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "cos_f",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 1, 512]",
+        "name" : "sin_f",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 512, 256]",
+        "name" : "kv13_k",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 256)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 512, 256]",
+        "name" : "kv13_v",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 2048, 512]",
+        "name" : "kv14_k",
+        "type" : "MultiArray"
+      },
+      {
+        "hasShapeFlexibility" : "0",
+        "isOptional" : "0",
+        "dataType" : "Float16",
+        "formattedType" : "MultiArray (Float16 1 × 1 × 2048 × 512)",
+        "shortDescription" : "",
+        "shape" : "[1, 1, 2048, 512]",
+        "name" : "kv14_v",
+        "type" : "MultiArray"
+      }
+    ],
+    "defaultFunctionName" : "infer",
+    "generatedClassName" : "chunk_3",
+    "userDefinedMetadata" : {
+
+    },
+    "method" : "predict"
+  }
+]

chunk_3.mlmodelc/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcf9130197bb9e56e106b559ef2cb767f54fef7eb5e621e8cd27111303ceac44
+size 527440000

cos_full.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4d3fc1ee45657134fda48f64d388bf70a68efe6d137abf2881ab4b6948af6d7
+size 8388736

cos_sliding.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a63098efaaf56e18c98f862987cf50202f0bb71d6f0def960fe1565260a7ffc
+size 4194432

embed_tokens_per_layer_q8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd1af367cde6fcb67a48925c0c566b0aac11946248d8e3cfebc4027550947cd8
+size 2348810240

embed_tokens_per_layer_scales.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a439606004e3372158eed3efd4c671cefea056122afb93c546108411ea41b057
+size 524288

embed_tokens_q8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2286db227b17e71fb3908983999876625c811295055457b6f798e0f09969797a
+size 402653184

embed_tokens_scales.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db47a8808d782d3c2b1ab722833b425e0bb4755360409d766acb89a4a8443461
+size 524288

hf_model/config.json

@@ -0,0 +1,190 @@
+{
+  "architectures": [
+    "Gemma4ForConditionalGeneration"
+  ],
+  "audio_config": {
+    "_name_or_path": "",
+    "architectures": null,
+    "attention_chunk_size": 12,
+    "attention_context_left": 13,
+    "attention_context_right": 0,
+    "attention_invalid_logits_value": -1000000000.0,
+    "attention_logit_cap": 50.0,
+    "chunk_size_feed_forward": 0,
+    "conv_kernel_size": 5,
+    "dtype": "bfloat16",
+    "gradient_clipping": 10000000000.0,
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
+    "initializer_range": 0.02,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
+    "model_type": "gemma4_audio",
+    "num_attention_heads": 8,
+    "num_hidden_layers": 12,
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "output_proj_dims": 1536,
+    "problem_type": null,
+    "residual_weight": 0.5,
+    "return_dict": true,
+    "rms_norm_eps": 1e-06,
+    "subsampling_conv_channels": [
+      128,
+      32
+    ],
+    "use_clipped_linears": true
+  },
+  "audio_token_id": 258881,
+  "boa_token_id": 256000,
+  "boi_token_id": 255999,
+  "dtype": "bfloat16",
+  "eoa_token_id": 258883,
+  "eoa_token_index": 258883,
+  "eoi_token_id": 258882,
+  "eos_token_id": [
+    1,
+    106
+  ],
+  "image_token_id": 258880,
+  "initializer_range": 0.02,
+  "model_type": "gemma4",
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attention_k_eq_v": false,
+    "bos_token_id": 2,
+    "dtype": "bfloat16",
+    "enable_moe_block": false,
+    "eos_token_id": 1,
+    "expert_intermediate_size": null,
+    "final_logit_softcapping": 30.0,
+    "global_head_dim": 512,
+    "head_dim": 256,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 1536,
+    "hidden_size_per_layer_input": 256,
+    "initializer_range": 0.02,
+    "intermediate_size": 6144,
+    "layer_types": [
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 131072,
+    "model_type": "gemma4_text",
+    "num_attention_heads": 8,
+    "num_experts": null,
+    "num_global_key_value_heads": null,
+    "num_hidden_layers": 35,
+    "num_key_value_heads": 1,
+    "num_kv_shared_layers": 20,
+    "pad_token_id": 0,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "full_attention": {
+        "partial_rotary_factor": 0.25,
+        "rope_theta": 1000000.0,
+        "rope_type": "proportional"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sliding_window": 512,
+    "tie_word_embeddings": true,
+    "top_k_experts": null,
+    "use_bidirectional_attention": null,
+    "use_cache": true,
+    "use_double_wide_mlp": true,
+    "vocab_size": 262144,
+    "vocab_size_per_layer_input": 262144
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.5.0.dev0",
+  "video_token_id": 258884,
+  "vision_config": {
+    "_name_or_path": "",
+    "architectures": null,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "chunk_size_feed_forward": 0,
+    "default_output_length": 280,
+    "dtype": "bfloat16",
+    "global_head_dim": 64,
+    "head_dim": 64,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 768,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
+    "initializer_range": 0.02,
+    "intermediate_size": 3072,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
+    "max_position_embeddings": 131072,
+    "model_type": "gemma4_vision",
+    "num_attention_heads": 12,
+    "num_hidden_layers": 16,
+    "num_key_value_heads": 12,
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "patch_size": 16,
+    "pooling_kernel_size": 3,
+    "position_embedding_size": 10240,
+    "problem_type": null,
+    "return_dict": true,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "rope_theta": 100.0,
+      "rope_type": "default"
+    },
+    "standardize": false,
+    "use_clipped_linears": true
+  },
+  "vision_soft_tokens_per_image": 280
+}

hf_model/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc8d3a0ce36466ccc1278bf987df5f71db1719b9ca6b4118264f45cb627bfe0f
+size 32169626

hf_model/tokenizer_config.json

@@ -0,0 +1,74 @@
+{
+  "audio_token": "<|audio|>",
+  "backend": "tokenizers",
+  "boa_token": "<|audio>",
+  "boi_token": "<|image>",
+  "bos_token": "<bos>",
+  "eoa_token": "<audio|>",
+  "eoc_token": "<channel|>",
+  "eoi_token": "<image|>",
+  "eos_token": "<eos>",
+  "eot_token": "<turn|>",
+  "escape_token": "<|\"|>",
+  "etc_token": "<tool_call|>",
+  "etd_token": "<tool|>",
+  "etr_token": "<tool_response|>",
+  "extra_special_tokens": [
+    "<|video|>"
+  ],
+  "image_token": "<|image|>",
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma4Processor",
+  "response_schema": {
+    "type": "object",
+    "properties": {
+      "role": {
+        "const": "assistant"
+      },
+      "thinking": {
+        "type": "string"
+      },
+      "content": {
+        "type": "string"
+      },
+      "tool_calls": {
+        "x-regex-iterator": "<\\|tool_call>(.*?)<tool_call\\|>",
+        "type": "array",
+        "items": {
+          "type": "object",
+          "properties": {
+            "type": {
+              "const": "function"
+            },
+            "function": {
+              "type": "object",
+              "x-regex": "call\\:(?P<name>\\w+)(?P<arguments>\\{.*\\})",
+              "properties": {
+                "name": {
+                  "type": "string"
+                },
+                "arguments": {
+                  "type": "object",
+                  "x-parser": "gemma4-tool-call",
+                  "additionalProperties": {}
+                }
+              }
+            }
+          }
+        }
+      }
+    },
+    "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<content>(?:(?!\\<\\|tool_call\\>)(?!\\<turn\\|\\>).)+)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?:\\<turn\\|\\>)?"
+  },
+  "soc_token": "<|channel>",
+  "sot_token": "<|turn>",
+  "stc_token": "<|tool_call>",
+  "std_token": "<|tool>",
+  "str_token": "<|tool_response>",
+  "think_token": "<|think|>",
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>"
+}

model_config.json

@@ -0,0 +1,23 @@
+{
+  "model_name": "gemma4-e2b-swa-ple",
+  "architecture": "gemma4",
+  "hidden_size": 1536,
+  "num_hidden_layers": 35,
+  "context_length": 2048,
+  "sliding_window": 512,
+  "vocab_size": 262144,
+  "bos_token_id": 2,
+  "eos_token_id": 1,
+  "per_layer_dim": 256,
+  "max_head_dim": 512,
+  "embed_scale": 39.191835884530846,
+  "per_layer_model_projection_scale": 0.02551551815399144,
+  "per_layer_input_scale": 0.7071067811865476,
+  "per_layer_embed_scale": 16.0,
+  "external_embeddings": true,
+  "has_multimodal": true,
+  "stateless": true,
+  "sliding_window_attention": true,
+  "ple_inside_chunk1": true,
+  "num_chunks": 4
+}

per_layer_norm_weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc90ef48ae4554761ff0c3a802ac0c6f08f4818fca7a7ecb7409a5a20da92a77
+size 1024

per_layer_projection.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d3db373cb38470e868fbf45bbddb13198fd61bbe57073340cfd9035028bb062
+size 27525120

sin_full.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79b947cd56c5043e1cae19d914dfc13051743229c95a6ed29a5ce03f4091e26f
+size 8388736

sin_sliding.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26fd591b8da63bf361713be334ec64dbd1d9809643fa8e74c359415d458f159c
+size 4194432