[ { "metadataOutputVersion" : "3.0", "storagePrecision" : "Float16", "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 4096 × 1 × 512)", "shortDescription" : "", "shape" : "[1, 4096, 1, 512]", "name" : "x", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 128 × 512)", "shortDescription" : "", "shape" : "[128, 512]", "name" : "cos", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 128 × 512)", "shortDescription" : "", "shape" : "[128, 512]", "name" : "sin", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 512)", "shortDescription" : "", "shape" : "[1, 1, 512, 512]", "name" : "mask", "type" : "MultiArray" } ], "modelParameters" : [ ], "specificationVersion" : 9, "functions" : [ { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1 × 512)", "shortDescription" : "", "shape" : "[1, 512]", "name" : "input_ids", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "full_sequence_length", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)", "storagePrecision" : "Float16", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 4096 × 1 × 512)", "shortDescription" : "", "shape" : "[1, 4096, 1, 512]", "name" : "x", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 128 × 512)", "shortDescription" : "", "shape" : "[128, 512]", "name" : "cos", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 128 × 512)", "shortDescription" : "", "shape" : "[128, 512]", "name" : "sin", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 512 × 512)", "shortDescription" : "", "shape" : "[1, 1, 512, 512]", "name" : "mask", "type" : "MultiArray" } ], "name" : "input_512_context_512", "mlProgramOperationTypeHistogram" : { "Select" : 2, "Ios18.maximum" : 1, "Ios18.gather" : 3, "Ios18.sub" : 3, "Ios18.transpose" : 1, "Ios18.less" : 2, "Ios18.cast" : 2, "Ios18.expandDims" : 4, "Tile" : 2 } }, { "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1 × 4)", "shortDescription" : "", "shape" : "[1, 4]", "name" : "input_ids", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "full_sequence_length", "type" : "MultiArray" } ], "computePrecision" : "Mixed (Float16, Int16, Int32, UInt16)", "storagePrecision" : "Float16", "stateSchema" : [ ], "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 4096 × 1 × 4)", "shortDescription" : "", "shape" : "[1, 4096, 1, 4]", "name" : "x", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 128 × 4)", "shortDescription" : "", "shape" : "[128, 4]", "name" : "cos", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 128 × 4)", "shortDescription" : "", "shape" : "[128, 4]", "name" : "sin", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 4 × 512)", "shortDescription" : "", "shape" : "[1, 1, 4, 512]", "name" : "mask", "type" : "MultiArray" } ], "name" : "input_1_context_512", "mlProgramOperationTypeHistogram" : { "Select" : 2, "Ios18.maximum" : 1, "Ios18.gather" : 3, "Ios18.sub" : 3, "Ios18.transpose" : 1, "Ios18.less" : 2, "Ios18.cast" : 2, "Ios18.expandDims" : 4, "Tile" : 2 } } ], "mlProgramOperationTypeHistogram" : { "Select" : 2, "Ios18.maximum" : 1, "Ios18.gather" : 3, "Ios18.sub" : 3, "Ios18.transpose" : 1, "Ios18.less" : 2, "Ios18.cast" : 2, "Ios18.expandDims" : 4, "Tile" : 2 }, "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, Int16, Int32, UInt16)", "modelType" : { "name" : "MLModelType_mlProgram" }, "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1 × 512)", "shortDescription" : "", "shape" : "[1, 512]", "name" : "input_ids", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "full_sequence_length", "type" : "MultiArray" } ], "defaultFunctionName" : "input_512_context_512", "generatedClassName" : "Llama_2_7b_hf_2024_07_17_19_34_17_merged_chunk1", "userDefinedMetadata" : { }, "method" : "predict" } ]