[ { "metadataOutputVersion" : "3.0", "storagePrecision" : "Mixed (Float16, Int32, Palettized (11 bits), Palettized (12 bits), Palettized (14 bits), UInt6)", "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 51865)", "shortDescription" : "", "shape" : "[1, 1, 51865]", "name" : "logits", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 24576 × 1 × 1)", "shortDescription" : "", "shape" : "[1, 24576, 1, 1]", "name" : "key_cache_updates", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 24576 × 1 × 1)", "shortDescription" : "", "shape" : "[1, 24576, 1, 1]", "name" : "value_cache_updates", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1500)", "shortDescription" : "", "shape" : "[1, 1500]", "name" : "alignment_heads_weights", "type" : "MultiArray" } ], "modelParameters" : [ ], "specificationVersion" : 9, "mlProgramOperationTypeHistogram" : { "Ios18.expandDims" : 6, "Ios18.softmax" : 48, "Ios18.mul" : 144, "Ios18.matmul" : 96, "Ios18.batchNorm" : 73, "Ios16.reduceMean" : 1, "Split" : 2, "Ios18.greaterEqual" : 2, "Select" : 2, "Ios18.gather" : 2, "Ios18.add" : 147, "Ios18.layerNorm" : 73, "Ios18.reshape" : 192, "Ios18.constexprLutToDense" : 241, "Ios18.conv" : 240, "Ios18.gelu" : 24, "Ios18.linear" : 1, "Ios18.sub" : 1, "Ios18.cast" : 1, "Ios18.transpose" : 1, "Ios18.concat" : 3, "Ios18.squeeze" : 1, "Ios18.sliceByIndex" : 12 }, "computePrecision" : "Mixed (Float16, Int32, UInt16)", "isUpdatable" : "0", "stateSchema" : [ ], "availability" : { "macOS" : "15.0", "tvOS" : "18.0", "visionOS" : "2.0", "watchOS" : "11.0", "iOS" : "18.0", "macCatalyst" : "18.0" }, "modelType" : { "name" : "MLModelType_mlProgram" }, "userDefinedMetadata" : { "com.github.apple.coremltools.conversion_date" : "2026-10-17", "com.github.apple.coremltools.source" : "torch==2.5.0", "com.github.apple.coremltools.version" : "9.0", "com.github.apple.coremltools.source_dialect" : "TorchScript" }, "inputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "input_ids", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Int32", "formattedType" : "MultiArray (Int32 1)", "shortDescription" : "", "shape" : "[1]", "name" : "cache_length", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 24576 × 1 × 448)", "shortDescription" : "", "shape" : "[1, 24576, 1, 448]", "name" : "key_cache", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 24576 × 1 × 448)", "shortDescription" : "", "shape" : "[1, 24576, 1, 448]", "name" : "value_cache", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 448)", "shortDescription" : "", "shape" : "[1, 448]", "name" : "kv_cache_update_mask", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 1500)", "shortDescription" : "", "shape" : "[1, 1024, 1, 1500]", "name" : "encoder_output_embeds", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 448)", "shortDescription" : "", "shape" : "[1, 448]", "name" : "decoder_key_padding_mask", "type" : "MultiArray" } ], "generatedClassName" : "TextDecoder_p6", "method" : "predict" } ]