PNG  IHDR;IDATxܻn0K )(pA 7LeG{ §㻢|ذaÆ 6lذaÆ 6lذaÆ 6lom$^yذag5bÆ 6lذaÆ 6lذa{ 6lذaÆ `}HFkm,mӪôô! x|'ܢ˟;E:9&ᶒ}{v]n&6 h_tڠ͵-ҫZ;Z$.Pkž)!o>}leQfJTu іچ\X=8Rن4`Vwl>nG^is"ms$ui?wbs[m6K4O.4%/bC%t Mז -lG6mrz2s%9s@-k9=)kB5\+͂Zsٲ Rn~GRC wIcIn7jJhۛNCS|j08yiHKֶۛkɈ+;SzL/F*\Ԕ#"5m2[S=gnaPeғL lذaÆ 6l^ḵaÆ 6lذaÆ 6lذa; _ذaÆ 6lذaÆ 6lذaÆ RIENDB` # Generated by default/object.tt package Paws::SageMaker::OutputConfig; use Moose; has CompilerOptions => (is => 'ro', isa => 'Str'); has KmsKeyId => (is => 'ro', isa => 'Str'); has S3OutputLocation => (is => 'ro', isa => 'Str', required => 1); has TargetDevice => (is => 'ro', isa => 'Str'); has TargetPlatform => (is => 'ro', isa => 'Paws::SageMaker::TargetPlatform'); 1; ### main pod documentation begin ### =head1 NAME Paws::SageMaker::OutputConfig =head1 USAGE This class represents one of two things: =head3 Arguments in a call to a service Use the attributes of this class as arguments to methods. You shouldn't make instances of this class. Each attribute should be used as a named argument in the calls that expect this type of object. As an example, if Att1 is expected to be a Paws::SageMaker::OutputConfig object: $service_obj->Method(Att1 => { CompilerOptions => $value, ..., TargetPlatform => $value }); =head3 Results returned from an API call Use accessors for each attribute. If Att1 is expected to be an Paws::SageMaker::OutputConfig object: $result = $service_obj->Method(...); $result->Att1->CompilerOptions =head1 DESCRIPTION Contains information about the output location for the compiled model and the target device that the model runs on. C and C are mutually exclusive, so you need to choose one between the two to specify your target device or platform. If you cannot find your device you want to use from the C list, use C to describe the platform of your edge device and C if there are specific settings that are required or recommended to use for particular TargetPlatform. =head1 ATTRIBUTES =head2 CompilerOptions => Str Specifies additional parameters for compiler options in JSON format. The compiler options are C specific. It is required for NVIDIA accelerators and highly recommended for CPU compilations. For any other cases, it is optional to specify C =over =item * C: Specifies the data type for the input. When compiling for C (except for C) instances using PyTorch framework, provide the data type (dtype) of the model's input. C<"float32"> is used if C<"DTYPE"> is not specified. Options for data type are: =over =item * float32: Use either C<"float"> or C<"float32">. =item * int64: Use either C<"int64"> or C<"long">. =back For example, C<{"dtype" : "float32"}>. =item * C: Compilation for CPU supports the following compiler options. =over =item * C: CPU micro-architecture. For example, C<{'mcpu': 'skylake-avx512'}> =item * C: CPU flags. For example, C<{'mattr': ['+neon', '+vfpv4']}> =back =item * C: Details of ARM CPU compilations. =over =item * C: NEON is an implementation of the Advanced SIMD extension used in ARMv7 processors. For example, add C<{'mattr': ['+neon']}> to the compiler options if compiling for ARM 32-bit platform with the NEON support. =back =item * C: Compilation for NVIDIA GPU supports the following compiler options. =over =item * C: Specifies the targeted architecture. =item * C: Specifies the TensorRT versions in x.y.z. format. =item * C: Specifies the CUDA version in x.y format. =back For example, C<{'gpu-code': 'sm_72', 'trt-ver': '6.0.1', 'cuda-ver': '10.1'}> =item * C: Compilation for the Android OS supports the following compiler options: =over =item * C: Specifies the Android API levels. Available levels range from 21 to 29. For example, C<{'ANDROID_PLATFORM': 28}>. =item * C: Add C<{'mattr': ['+neon']}> to compiler options if compiling for ARM 32-bit platform with NEON support. =back =item * C: Compilation for target ml_inf1 uses compiler options passed in as a JSON string. For example, C<"CompilerOptions": "\"--verbose 1 --num-neuroncores 2 -O2\"">. For information about supported compiler options, see Neuron Compiler CLI (https://github.com/aws/aws-neuron-sdk/blob/master/docs/neuron-cc/command-line-reference.md). =item * C: Compilation for the CoreML OutputConfig$TargetDevice supports the following compiler options: =over =item * C: Specifies the classification labels file name inside input tar.gz file. For example, C<{"class_labels": "imagenet_labels_1000.txt"}>. Labels inside the txt file should be separated by newlines. =back =item * C: Compilation for the Elastic Inference Accelerator supports the following compiler options: =over =item * C: Specifies the precision of compiled artifacts. Supported values are C<"FP16"> and C<"FP32">. Default is C<"FP32">. =item * C: Specifies the signature to use for models in SavedModel format. Defaults is TensorFlow's default signature def key. =item * C: Specifies a list of output tensor names for models in FrozenGraph format. Set at most one API field, either: C or C. =back For example: C<{"precision_mode": "FP32", "output_names": ["output:0"]}> =back =head2 KmsKeyId => Str The Amazon Web Services Key Management Service key (Amazon Web Services KMS) that Amazon SageMaker uses to encrypt your output models with Amazon S3 server-side encryption after compilation job. If you don't provide a KMS key ID, Amazon SageMaker uses the default KMS key for Amazon S3 for your role's account. For more information, see KMS-Managed Encryption Keys (https://docs.aws.amazon.com/AmazonS3/latest/userguide/UsingKMSEncryption.html) in the I The KmsKeyId can be any of the following formats: =over =item * Key ID: C<1234abcd-12ab-34cd-56ef-1234567890ab> =item * Key ARN: C =item * Alias name: C =item * Alias name ARN: C =back =head2 B S3OutputLocation => Str Identifies the S3 bucket where you want Amazon SageMaker to store the model artifacts. For example, C. =head2 TargetDevice => Str Identifies the target device or the machine learning instance that you want to run your model on after the compilation has completed. Alternatively, you can specify OS, architecture, and accelerator using TargetPlatform fields. It can be used instead of C. =head2 TargetPlatform => L Contains information about a target platform that you want your model to run on, such as OS, architecture, and accelerators. It is an alternative of C. The following examples show how to configure the C and C JSON strings for popular target platforms: =over =item * Raspberry Pi 3 Model B+ C<"TargetPlatform": {"Os": "LINUX", "Arch": "ARM_EABIHF"},> C<"CompilerOptions": {'mattr': ['+neon']}> =item * Jetson TX2 C<"TargetPlatform": {"Os": "LINUX", "Arch": "ARM64", "Accelerator": "NVIDIA"},> C<"CompilerOptions": {'gpu-code': 'sm_62', 'trt-ver': '6.0.1', 'cuda-ver': '10.0'}> =item * EC2 m5.2xlarge instance OS C<"TargetPlatform": {"Os": "LINUX", "Arch": "X86_64", "Accelerator": "NVIDIA"},> C<"CompilerOptions": {'mcpu': 'skylake-avx512'}> =item * RK3399 C<"TargetPlatform": {"Os": "LINUX", "Arch": "ARM64", "Accelerator": "MALI"}> =item * ARMv7 phone (CPU) C<"TargetPlatform": {"Os": "ANDROID", "Arch": "ARM_EABI"},> C<"CompilerOptions": {'ANDROID_PLATFORM': 25, 'mattr': ['+neon']}> =item * ARMv8 phone (CPU) C<"TargetPlatform": {"Os": "ANDROID", "Arch": "ARM64"},> C<"CompilerOptions": {'ANDROID_PLATFORM': 29}> =back =head1 SEE ALSO This class forms part of L, describing an object used in L =head1 BUGS and CONTRIBUTIONS The source code is located here: L Please report bugs to: L =cut