[ { "metadataOutputVersion" : "3.0", "storagePrecision" : "Mixed (Float16, Palettized (4 bits), Palettized (6 bits))", "outputSchema" : [ { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1 × 51866)", "shortDescription" : "", "shape" : "[1, 1, 51866]", "name" : "logits", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 40960 × 1 × 1)", "shortDescription" : "", "shape" : "[1, 40960, 1, 1]", "name" : "key_cache_updates", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 40960 × 1 × 1)", "shortDescription" : "", "shape" : "[1, 40960, 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" : 8, "mlProgramOperationTypeHistogram" : { "Ios17.reshape" : 256, "Ios16.reduceMean" : 195, "Ios16.softmax" : 64, "Ios17.matmul" : 128, "Ios17.transpose" : 1, "Split" : 2, "Ios17.expandDims" : 6, "Ios17.add" : 610, "Ios17.sliceByIndex" : 20, "Ios17.squeeze" : 1, "Ios17.gather" : 2, "Ios17.batchNorm" : 97, "Ios16.gelu" : 32, "Ios17.cast" : 1, "Ios16.constexprLutToDense" : 320, "Ios17.conv" : 960, "Ios17.sub" : 98, "Ios17.concat" : 3, "Ios17.linear" : 1, "Ios17.rsqrt" : 97, "Ios17.mul" : 386 }, "computePrecision" : "Mixed (Float16, Float32, Int16, Int32)", "isUpdatable" : "0", "availability" : { "macOS" : "14.0", "tvOS" : "17.0", "visionOS" : "1.0", "watchOS" : "10.0", "iOS" : "17.0", "macCatalyst" : "17.0" }, "modelType" : { "name" : "MLModelType_mlProgram" }, "userDefinedMetadata" : { "com.github.apple.coremltools.source_dialect" : "TorchScript", "com.github.apple.coremltools.version" : "7.1", "com.github.apple.coremltools.source" : "torch==2.2.1" }, "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 × 40960 × 1 × 224)", "shortDescription" : "", "shape" : "[1, 40960, 1, 224]", "name" : "key_cache", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 40960 × 1 × 224)", "shortDescription" : "", "shape" : "[1, 40960, 1, 224]", "name" : "value_cache", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 224)", "shortDescription" : "", "shape" : "[1, 224]", "name" : "kv_cache_update_mask", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 1280 × 1 × 1500)", "shortDescription" : "", "shape" : "[1, 1280, 1, 1500]", "name" : "encoder_output_embeds", "type" : "MultiArray" }, { "hasShapeFlexibility" : "0", "isOptional" : "0", "dataType" : "Float16", "formattedType" : "MultiArray (Float16 1 × 224)", "shortDescription" : "", "shape" : "[1, 224]", "name" : "decoder_key_padding_mask", "type" : "MultiArray" } ], "generatedClassName" : "TextDecoder_mixedBitPalettized_4_0_bit", "method" : "predict" } ]