[ { "metadataOutputVersion" : "3.0", "storagePrecision" : "Mixed (Float16, Int32, Int8)", "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 1024 × 63)", "shortDescription" : "", "shape" : "[1, 1024, 63]", "name" : "encoded", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "encoded_length", "type" : "MultiArray" } ], "modelParameters" : [ ], "specificationVersion" : 9, "functions" : [ { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 128 × 501)", "shortDescription" : "", "shape" : "[1, 128, 501]", "name" : "audio_signal", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "length", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Float32, Int32)", "storagePrecision" : "Mixed (Float16, Int32, Int8)", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 1024 × 63)", "shortDescription" : "", "shape" : "[1, 1024, 63]", "name" : "encoded", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "encoded_length", "type" : "MultiArray" } ], "name" : "bucket_501", "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 14, "Ios18.mul" : 128, "Ios18.softmax" : 24, "Ios18.floorDiv" : 4, "Ios18.matmul" : 72, "Ios18.batchNorm" : 24, "Ios18.logicalNot" : 1, "Ios18.floor" : 3, "Ios18.equal" : 24, "Split" : 24, "Select" : 96, "Ios18.sigmoid" : 24, "Tile" : 1, "Ios18.add" : 184, "Ios16.reduceSum" : 1, "Ios18.layerNorm" : 120, "Ios16.reduceMin" : 1, "Ios18.reshape" : 145, "Pad" : 24, "Ios18.linear" : 193, "Ios18.relu" : 3, "Ios18.conv" : 77, "Ios18.sub" : 3, "Ios18.logicalAnd" : 1, "Ios18.silu" : 72, "Ios18.transpose" : 148, "Ios18.cast" : 9, "Ios18.less" : 5, "Ios18.sliceByIndex" : 48, "Ios18.constexprBlockwiseShiftScale" : 270 } }, { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 128 × 1501)", "shortDescription" : "", "shape" : "[1, 128, 1501]", "name" : "audio_signal", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "length", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Float32, Int32)", "storagePrecision" : "Mixed (Float16, Int32, Int8)", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 1024 × 188)", "shortDescription" : "", "shape" : "[1, 1024, 188]", "name" : "encoded", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "encoded_length", "type" : "MultiArray" } ], "name" : "bucket_1501", "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 14, "Ios18.mul" : 128, "Ios18.softmax" : 24, "Ios18.floorDiv" : 4, "Ios18.matmul" : 72, "Ios18.batchNorm" : 24, "Ios18.logicalNot" : 1, "Ios18.floor" : 3, "Ios18.equal" : 24, "Split" : 24, "Select" : 96, "Ios18.sigmoid" : 24, "Tile" : 1, "Ios18.add" : 184, "Ios16.reduceSum" : 1, "Ios18.layerNorm" : 120, "Ios16.reduceMin" : 1, "Ios18.reshape" : 145, "Pad" : 24, "Ios18.linear" : 193, "Ios18.relu" : 3, "Ios18.conv" : 77, "Ios18.sub" : 3, "Ios18.logicalAnd" : 1, "Ios18.silu" : 72, "Ios18.transpose" : 148, "Ios18.cast" : 9, "Ios18.less" : 5, "Ios18.sliceByIndex" : 48, "Ios18.constexprBlockwiseShiftScale" : 270 } }, { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 128 × 3001)", "shortDescription" : "", "shape" : "[1, 128, 3001]", "name" : "audio_signal", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "length", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Float32, Int32)", "storagePrecision" : "Mixed (Float16, Int32, Int8)", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 1024 × 376)", "shortDescription" : "", "shape" : "[1, 1024, 376]", "name" : "encoded", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "encoded_length", "type" : "MultiArray" } ], "name" : "bucket_3001", "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 14, "Ios18.mul" : 128, "Ios18.softmax" : 24, "Ios18.floorDiv" : 4, "Ios18.matmul" : 72, "Ios18.batchNorm" : 24, "Ios18.logicalNot" : 1, "Ios18.floor" : 3, "Ios18.equal" : 24, "Split" : 24, "Select" : 96, "Ios18.sigmoid" : 24, "Tile" : 1, "Ios18.add" : 184, "Ios16.reduceSum" : 1, "Ios18.layerNorm" : 120, "Ios16.reduceMin" : 1, "Ios18.reshape" : 145, "Pad" : 24, "Ios18.linear" : 193, "Ios18.relu" : 3, "Ios18.conv" : 77, "Ios18.sub" : 3, "Ios18.logicalAnd" : 1, "Ios18.silu" : 72, "Ios18.transpose" : 148, "Ios18.cast" : 9, "Ios18.less" : 5, "Ios18.sliceByIndex" : 48, "Ios18.constexprBlockwiseShiftScale" : 270 } } ], "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 14, "Ios18.mul" : 128, "Ios18.softmax" : 24, "Ios18.floorDiv" : 4, "Ios18.matmul" : 72, "Ios18.batchNorm" : 24, "Ios18.logicalNot" : 1, "Ios18.floor" : 3, "Ios18.equal" : 24, "Split" : 24, "Select" : 96, "Ios18.sigmoid" : 24, "Tile" : 1, "Ios18.add" : 184, "Ios16.reduceSum" : 1, "Ios18.layerNorm" : 120, "Ios16.reduceMin" : 1, "Ios18.reshape" : 145, "Pad" : 24, "Ios18.linear" : 193, "Ios18.relu" : 3, "Ios18.conv" : 77, "Ios18.sub" : 3, "Ios18.logicalAnd" : 1, "Ios18.silu" : 72, "Ios18.transpose" : 148, "Ios18.cast" : 9, "Ios18.less" : 5, "Ios18.sliceByIndex" : 48, "Ios18.constexprBlockwiseShiftScale" : 270 }, "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, Float32, Int32)", "modelType" : { "name" : "MLModelType_mlProgram" }, "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float32", "formattedType" : "MultiArray (Float32 1 × 128 × 501)", "shortDescription" : "", "shape" : "[1, 128, 501]", "name" : "audio_signal", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "length", "type" : "MultiArray" } ], "defaultFunctionName" : "bucket_501", "generatedClassName" : "encoder", "userDefinedMetadata" : { }, "method" : "predict" } ]