[ { "metadataOutputVersion" : "3.0", "storagePrecision" : "Float16", "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 12 × 2048)", "shortDescription" : "", "shape" : "[1, 12, 2048]", "name" : "emb", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v_out", "type" : "MultiArray" } ], "modelParameters" : [ ], "specificationVersion" : 9, "functions" : [ { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 128 × 104)", "shortDescription" : "", "shape" : "[1, 128, 104]", "name" : "mel", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 52)", "shortDescription" : "", "shape" : "[1, 1, 1, 52]", "name" : "m1", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 26)", "shortDescription" : "", "shape" : "[1, 1, 1, 26]", "name" : "m2", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 12 × 43)", "shortDescription" : "", "shape" : "[1, 1, 12, 43]", "name" : "mask", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Float32, Int32)", "storagePrecision" : "Float16", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 12 × 2048)", "shortDescription" : "", "shape" : "[1, 12, 2048]", "name" : "emb", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v_out", "type" : "MultiArray" } ], "name" : "n12", "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 15, "Ios18.mul" : 275, "Ios18.rsqrt" : 49, "Ios18.matmul" : 14, "Ios16.reduceMean" : 49, "Ios18.realDiv" : 105, "Ios18.exp" : 7, "Split" : 14, "Ios16.reduceMax" : 56, "Tile" : 14, "Ios18.add" : 148, "Ios16.reduceSum" : 7, "Ios18.layerNorm" : 1, "Ios18.reshape" : 43, "Pad" : 48, "Ios18.inverse" : 49, "Ios18.maximum" : 49, "Ios18.linear" : 101, "Ios18.conv" : 16, "Ios18.concat" : 14, "Ios18.tanh" : 20, "Ios18.transpose" : 49, "Ios18.pow" : 49, "Ios18.sub" : 7, "Ios18.abs" : 49, "Stack" : 3, "Ios18.gelu" : 1, "Ios18.sliceByIndex" : 79 } }, { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 128 × 776)", "shortDescription" : "", "shape" : "[1, 128, 776]", "name" : "mel", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 388)", "shortDescription" : "", "shape" : "[1, 1, 1, 388]", "name" : "m1", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 194)", "shortDescription" : "", "shape" : "[1, 1, 1, 194]", "name" : "m2", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 96 × 127)", "shortDescription" : "", "shape" : "[1, 1, 96, 127]", "name" : "mask", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Float32, Int32)", "storagePrecision" : "Float16", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 96 × 2048)", "shortDescription" : "", "shape" : "[1, 96, 2048]", "name" : "emb", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v_out", "type" : "MultiArray" } ], "name" : "n96", "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 15, "Ios18.mul" : 275, "Ios18.rsqrt" : 49, "Ios18.matmul" : 14, "Ios16.reduceMean" : 49, "Ios18.realDiv" : 105, "Ios18.exp" : 7, "Split" : 14, "Ios16.reduceMax" : 56, "Tile" : 14, "Ios18.add" : 148, "Ios16.reduceSum" : 7, "Ios18.layerNorm" : 1, "Ios18.reshape" : 43, "Pad" : 48, "Ios18.inverse" : 49, "Ios18.maximum" : 49, "Ios18.linear" : 101, "Ios18.conv" : 16, "Ios18.concat" : 14, "Ios18.tanh" : 20, "Ios18.transpose" : 49, "Ios18.pow" : 49, "Ios18.sub" : 7, "Ios18.abs" : 49, "Stack" : 3, "Ios18.gelu" : 1, "Ios18.sliceByIndex" : 79 } }, { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 128 × 3008)", "shortDescription" : "", "shape" : "[1, 128, 3008]", "name" : "mel", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 1504)", "shortDescription" : "", "shape" : "[1, 1, 1, 1504]", "name" : "m1", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 752)", "shortDescription" : "", "shape" : "[1, 1, 1, 752]", "name" : "m2", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 375 × 406)", "shortDescription" : "", "shape" : "[1, 1, 375, 406]", "name" : "mask", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Float32, Int32)", "storagePrecision" : "Float16", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 375 × 2048)", "shortDescription" : "", "shape" : "[1, 375, 2048]", "name" : "emb", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k_out", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v_out", "type" : "MultiArray" } ], "name" : "n375", "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 15, "Ios18.mul" : 275, "Ios18.rsqrt" : 49, "Ios18.matmul" : 14, "Ios16.reduceMean" : 49, "Ios18.realDiv" : 105, "Ios18.exp" : 7, "Split" : 14, "Ios16.reduceMax" : 56, "Tile" : 14, "Ios18.add" : 148, "Ios16.reduceSum" : 7, "Ios18.layerNorm" : 1, "Ios18.reshape" : 43, "Pad" : 48, "Ios18.inverse" : 49, "Ios18.maximum" : 49, "Ios18.linear" : 101, "Ios18.conv" : 16, "Ios18.concat" : 14, "Ios18.tanh" : 20, "Ios18.transpose" : 49, "Ios18.pow" : 49, "Ios18.sub" : 7, "Ios18.abs" : 49, "Stack" : 3, "Ios18.gelu" : 1, "Ios18.sliceByIndex" : 79 } } ], "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 15, "Ios18.mul" : 275, "Ios18.rsqrt" : 49, "Ios18.matmul" : 14, "Ios16.reduceMean" : 49, "Ios18.realDiv" : 105, "Ios18.exp" : 7, "Split" : 14, "Ios16.reduceMax" : 56, "Tile" : 14, "Ios18.add" : 148, "Ios16.reduceSum" : 7, "Ios18.layerNorm" : 1, "Ios18.reshape" : 43, "Pad" : 48, "Ios18.inverse" : 49, "Ios18.maximum" : 49, "Ios18.linear" : 101, "Ios18.conv" : 16, "Ios18.concat" : 14, "Ios18.tanh" : 20, "Ios18.transpose" : 49, "Ios18.pow" : 49, "Ios18.sub" : 7, "Ios18.abs" : 49, "Stack" : 3, "Ios18.gelu" : 1, "Ios18.sliceByIndex" : 79 }, "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" : "Float16", "formattedType" : "MultiArray (Float16 1 × 128 × 104)", "shortDescription" : "", "shape" : "[1, 128, 104]", "name" : "mel", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 52)", "shortDescription" : "", "shape" : "[1, 1, 1, 52]", "name" : "m1", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 26)", "shortDescription" : "", "shape" : "[1, 1, 1, 26]", "name" : "m2", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 10 × 1 × 768 × 2)", "shortDescription" : "", "shape" : "[10, 1, 768, 2]", "name" : "conv", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "k", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 7 × 1 × 4 × 31 × 96)", "shortDescription" : "", "shape" : "[7, 1, 4, 31, 96]", "name" : "v", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 12 × 43)", "shortDescription" : "", "shape" : "[1, 1, 12, 43]", "name" : "mask", "type" : "MultiArray" } ], "defaultFunctionName" : "n12", "generatedClassName" : "AudioEncoder_fp16", "userDefinedMetadata" : { }, "method" : "predict" } ]