[
  {
    "metadataOutputVersion" : "3.0",
    "storagePrecision" : "Mixed (Float16, Int8)",
    "outputSchema" : [
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Int32",
        "formattedType" : "MultiArray (Int32 1 × 1)",
        "shortDescription" : "",
        "shape" : "[1, 1]",
        "name" : "nextToken",
        "type" : "MultiArray"
      },
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Float16",
        "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
        "shortDescription" : "",
        "shape" : "[28, 8, 16, 128]",
        "name" : "newKeys",
        "type" : "MultiArray"
      },
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Float16",
        "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
        "shortDescription" : "",
        "shape" : "[28, 8, 16, 128]",
        "name" : "newValues",
        "type" : "MultiArray"
      }
    ],
    "modelParameters" : [

    ],
    "specificationVersion" : 9,
    "functions" : [
      {
        "inputSchema" : [
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1 × 16)",
            "shortDescription" : "",
            "shape" : "[1, 16]",
            "name" : "inputIds",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 16)",
            "shortDescription" : "",
            "shape" : "[16]",
            "name" : "positions",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1)",
            "shortDescription" : "",
            "shape" : "[1]",
            "name" : "answerRow",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
            "shortDescription" : "",
            "shape" : "[1, 1, 1, 512]",
            "name" : "cacheMask",
            "type" : "MultiArray"
          }
        ],
        "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
        "storagePrecision" : "Mixed (Float16, Int8)",
        "stateSchema" : [
          {
            "dataType" : "Float16",
            "isOptional" : "0",
            "formattedType" : "State (Float16 28 × 8 × 528 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 528, 128]",
            "name" : "keyCache",
            "type" : "State"
          },
          {
            "dataType" : "Float16",
            "isOptional" : "0",
            "formattedType" : "State (Float16 28 × 8 × 528 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 528, 128]",
            "name" : "valueCache",
            "type" : "State"
          }
        ],
        "outputSchema" : [
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1 × 1)",
            "shortDescription" : "",
            "shape" : "[1, 1]",
            "name" : "nextToken",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 16, 128]",
            "name" : "newKeys",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 16, 128]",
            "name" : "newValues",
            "type" : "MultiArray"
          }
        ],
        "name" : "window512",
        "mlProgramOperationTypeHistogram" : {
          "Ios18.softmax" : 28,
          "Ios18.mul" : 534,
          "Ios18.matmul" : 112,
          "Ios18.rsqrt" : 113,
          "Ios16.reduceMean" : 113,
          "Ios18.greaterEqual" : 2,
          "Select" : 2,
          "Ios18.readState" : 2,
          "Ios18.gather" : 4,
          "Ios18.add" : 311,
          "Ios18.reduceArgmax" : 1,
          "Ios18.reshape" : 114,
          "Ios18.linear" : 197,
          "Ios18.concat" : 142,
          "Ios18.cast" : 2,
          "Ios18.sub" : 56,
          "Ios18.pow" : 113,
          "Ios18.transpose" : 140,
          "Ios18.silu" : 28,
          "Ios18.sliceByIndex" : 392,
          "Ios18.constexprBlockwiseShiftScale" : 198
        }
      },
      {
        "inputSchema" : [
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1 × 16)",
            "shortDescription" : "",
            "shape" : "[1, 16]",
            "name" : "inputIds",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 16)",
            "shortDescription" : "",
            "shape" : "[16]",
            "name" : "positions",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1)",
            "shortDescription" : "",
            "shape" : "[1]",
            "name" : "answerRow",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 2048)",
            "shortDescription" : "",
            "shape" : "[1, 1, 1, 2048]",
            "name" : "cacheMask",
            "type" : "MultiArray"
          }
        ],
        "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
        "storagePrecision" : "Mixed (Float16, Int8)",
        "stateSchema" : [
          {
            "dataType" : "Float16",
            "isOptional" : "0",
            "formattedType" : "State (Float16 28 × 8 × 2064 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 2064, 128]",
            "name" : "keyCache",
            "type" : "State"
          },
          {
            "dataType" : "Float16",
            "isOptional" : "0",
            "formattedType" : "State (Float16 28 × 8 × 2064 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 2064, 128]",
            "name" : "valueCache",
            "type" : "State"
          }
        ],
        "outputSchema" : [
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1 × 1)",
            "shortDescription" : "",
            "shape" : "[1, 1]",
            "name" : "nextToken",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 16, 128]",
            "name" : "newKeys",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 16, 128]",
            "name" : "newValues",
            "type" : "MultiArray"
          }
        ],
        "name" : "window2048",
        "mlProgramOperationTypeHistogram" : {
          "Ios18.softmax" : 28,
          "Ios18.mul" : 534,
          "Ios18.matmul" : 112,
          "Ios18.rsqrt" : 113,
          "Ios16.reduceMean" : 113,
          "Ios18.greaterEqual" : 2,
          "Select" : 2,
          "Ios18.readState" : 2,
          "Ios18.gather" : 4,
          "Ios18.add" : 311,
          "Ios18.reduceArgmax" : 1,
          "Ios18.reshape" : 114,
          "Ios18.linear" : 197,
          "Ios18.concat" : 142,
          "Ios18.cast" : 2,
          "Ios18.sub" : 56,
          "Ios18.pow" : 113,
          "Ios18.transpose" : 140,
          "Ios18.silu" : 28,
          "Ios18.sliceByIndex" : 392,
          "Ios18.constexprBlockwiseShiftScale" : 198
        }
      },
      {
        "inputSchema" : [
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1 × 16)",
            "shortDescription" : "",
            "shape" : "[1, 16]",
            "name" : "inputIds",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 16)",
            "shortDescription" : "",
            "shape" : "[16]",
            "name" : "positions",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1)",
            "shortDescription" : "",
            "shape" : "[1]",
            "name" : "answerRow",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 4096)",
            "shortDescription" : "",
            "shape" : "[1, 1, 1, 4096]",
            "name" : "cacheMask",
            "type" : "MultiArray"
          }
        ],
        "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
        "storagePrecision" : "Mixed (Float16, Int8)",
        "stateSchema" : [
          {
            "dataType" : "Float16",
            "isOptional" : "0",
            "formattedType" : "State (Float16 28 × 8 × 4112 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 4112, 128]",
            "name" : "keyCache",
            "type" : "State"
          },
          {
            "dataType" : "Float16",
            "isOptional" : "0",
            "formattedType" : "State (Float16 28 × 8 × 4112 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 4112, 128]",
            "name" : "valueCache",
            "type" : "State"
          }
        ],
        "outputSchema" : [
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Int32",
            "formattedType" : "MultiArray (Int32 1 × 1)",
            "shortDescription" : "",
            "shape" : "[1, 1]",
            "name" : "nextToken",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 16, 128]",
            "name" : "newKeys",
            "type" : "MultiArray"
          },
          {
            "hasShapeFlexibility" : "0",
            "isOptional" : "0",
            "dataType" : "Float16",
            "formattedType" : "MultiArray (Float16 28 × 8 × 16 × 128)",
            "shortDescription" : "",
            "shape" : "[28, 8, 16, 128]",
            "name" : "newValues",
            "type" : "MultiArray"
          }
        ],
        "name" : "window4096",
        "mlProgramOperationTypeHistogram" : {
          "Ios18.softmax" : 28,
          "Ios18.mul" : 534,
          "Ios18.matmul" : 112,
          "Ios18.rsqrt" : 113,
          "Ios16.reduceMean" : 113,
          "Ios18.greaterEqual" : 2,
          "Select" : 2,
          "Ios18.readState" : 2,
          "Ios18.gather" : 4,
          "Ios18.add" : 311,
          "Ios18.reduceArgmax" : 1,
          "Ios18.reshape" : 114,
          "Ios18.linear" : 197,
          "Ios18.concat" : 142,
          "Ios18.cast" : 2,
          "Ios18.sub" : 56,
          "Ios18.pow" : 113,
          "Ios18.transpose" : 140,
          "Ios18.silu" : 28,
          "Ios18.sliceByIndex" : 392,
          "Ios18.constexprBlockwiseShiftScale" : 198
        }
      }
    ],
    "mlProgramOperationTypeHistogram" : {
      "Ios18.softmax" : 28,
      "Ios18.mul" : 534,
      "Ios18.matmul" : 112,
      "Ios18.rsqrt" : 113,
      "Ios16.reduceMean" : 113,
      "Ios18.greaterEqual" : 2,
      "Select" : 2,
      "Ios18.readState" : 2,
      "Ios18.gather" : 4,
      "Ios18.add" : 311,
      "Ios18.reduceArgmax" : 1,
      "Ios18.reshape" : 114,
      "Ios18.linear" : 197,
      "Ios18.concat" : 142,
      "Ios18.cast" : 2,
      "Ios18.sub" : 56,
      "Ios18.pow" : 113,
      "Ios18.transpose" : 140,
      "Ios18.silu" : 28,
      "Ios18.sliceByIndex" : 392,
      "Ios18.constexprBlockwiseShiftScale" : 198
    },
    "isUpdatable" : "0",
    "stateSchema" : [
      {
        "dataType" : "Float16",
        "isOptional" : "0",
        "formattedType" : "State (Float16 28 × 8 × 528 × 128)",
        "shortDescription" : "",
        "shape" : "[28, 8, 528, 128]",
        "name" : "keyCache",
        "type" : "State"
      },
      {
        "dataType" : "Float16",
        "isOptional" : "0",
        "formattedType" : "State (Float16 28 × 8 × 528 × 128)",
        "shortDescription" : "",
        "shape" : "[28, 8, 528, 128]",
        "name" : "valueCache",
        "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, UInt16)",
    "modelType" : {
      "name" : "MLModelType_mlProgram"
    },
    "inputSchema" : [
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Int32",
        "formattedType" : "MultiArray (Int32 1 × 16)",
        "shortDescription" : "",
        "shape" : "[1, 16]",
        "name" : "inputIds",
        "type" : "MultiArray"
      },
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Int32",
        "formattedType" : "MultiArray (Int32 16)",
        "shortDescription" : "",
        "shape" : "[16]",
        "name" : "positions",
        "type" : "MultiArray"
      },
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Int32",
        "formattedType" : "MultiArray (Int32 1)",
        "shortDescription" : "",
        "shape" : "[1]",
        "name" : "answerRow",
        "type" : "MultiArray"
      },
      {
        "hasShapeFlexibility" : "0",
        "isOptional" : "0",
        "dataType" : "Float16",
        "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 512)",
        "shortDescription" : "",
        "shape" : "[1, 1, 1, 512]",
        "name" : "cacheMask",
        "type" : "MultiArray"
      }
    ],
    "defaultFunctionName" : "window512",
    "generatedClassName" : "CleanupNeuralEngine",
    "userDefinedMetadata" : {

    },
    "method" : "predict"
  }
]