| 1 | # Generated by default/object.tt | ||
| 2 | package Paws::SageMaker::AlgorithmSpecification; | ||
| 3 | use Moose; | ||
| 4 | has AlgorithmName => (is => 'ro', isa => 'Str'); | ||
| 5 | has EnableSageMakerMetricsTimeSeries => (is => 'ro', isa => 'Bool'); | ||
| 6 | has MetricDefinitions => (is => 'ro', isa => 'ArrayRef[Paws::SageMaker::MetricDefinition]'); | ||
| 7 | has TrainingImage => (is => 'ro', isa => 'Str'); | ||
| 8 | has TrainingInputMode => (is => 'ro', isa => 'Str', required => 1); | ||
| 9 | |||
| 10 | 1; | ||
| 11 | |||
| 12 | ### main pod documentation begin ### | ||
| 13 | |||
| 14 | =head1 NAME | ||
| 15 | |||
| 16 | Paws::SageMaker::AlgorithmSpecification | ||
| 17 | |||
| 18 | =head1 USAGE | ||
| 19 | |||
| 20 | This class represents one of two things: | ||
| 21 | |||
| 22 | =head3 Arguments in a call to a service | ||
| 23 | |||
| 24 | Use the attributes of this class as arguments to methods. You shouldn't make instances of this class. | ||
| 25 | Each attribute should be used as a named argument in the calls that expect this type of object. | ||
| 26 | |||
| 27 | As an example, if Att1 is expected to be a Paws::SageMaker::AlgorithmSpecification object: | ||
| 28 | |||
| 29 | $service_obj->Method(Att1 => { AlgorithmName => $value, ..., TrainingInputMode => $value }); | ||
| 30 | |||
| 31 | =head3 Results returned from an API call | ||
| 32 | |||
| 33 | Use accessors for each attribute. If Att1 is expected to be an Paws::SageMaker::AlgorithmSpecification object: | ||
| 34 | |||
| 35 | $result = $service_obj->Method(...); | ||
| 36 | $result->Att1->AlgorithmName | ||
| 37 | |||
| 38 | =head1 DESCRIPTION | ||
| 39 | |||
| 40 | Specifies the training algorithm to use in a CreateTrainingJob request. | ||
| 41 | |||
| 42 | For more information about algorithms provided by Amazon SageMaker, see | ||
| 43 | Algorithms | ||
| 44 | (https://docs.aws.amazon.com/sagemaker/latest/dg/algos.html). For | ||
| 45 | information about using your own algorithms, see Using Your Own | ||
| 46 | Algorithms with Amazon SageMaker | ||
| 47 | (https://docs.aws.amazon.com/sagemaker/latest/dg/your-algorithms.html). | ||
| 48 | |||
| 49 | =head1 ATTRIBUTES | ||
| 50 | |||
| 51 | |||
| 52 | =head2 AlgorithmName => Str | ||
| 53 | |||
| 54 | The name of the algorithm resource to use for the training job. This | ||
| 55 | must be an algorithm resource that you created or subscribe to on | ||
| 56 | Amazon Web Services Marketplace. If you specify a value for this | ||
| 57 | parameter, you can't specify a value for C<TrainingImage>. | ||
| 58 | |||
| 59 | |||
| 60 | =head2 EnableSageMakerMetricsTimeSeries => Bool | ||
| 61 | |||
| 62 | To generate and save time-series metrics during training, set to | ||
| 63 | C<true>. The default is C<false> and time-series metrics aren't | ||
| 64 | generated except in the following cases: | ||
| 65 | |||
| 66 | =over | ||
| 67 | |||
| 68 | =item * | ||
| 69 | |||
| 70 | You use one of the Amazon SageMaker built-in algorithms | ||
| 71 | |||
| 72 | =item * | ||
| 73 | |||
| 74 | You use one of the following Prebuilt Amazon SageMaker Docker Images | ||
| 75 | (https://docs.aws.amazon.com/sagemaker/latest/dg/pre-built-containers-frameworks-deep-learning.html): | ||
| 76 | |||
| 77 | =over | ||
| 78 | |||
| 79 | =item * | ||
| 80 | |||
| 81 | Tensorflow (version E<gt>= 1.15) | ||
| 82 | |||
| 83 | =item * | ||
| 84 | |||
| 85 | MXNet (version E<gt>= 1.6) | ||
| 86 | |||
| 87 | =item * | ||
| 88 | |||
| 89 | PyTorch (version E<gt>= 1.3) | ||
| 90 | |||
| 91 | =back | ||
| 92 | |||
| 93 | =item * | ||
| 94 | |||
| 95 | You specify at least one MetricDefinition | ||
| 96 | |||
| 97 | =back | ||
| 98 | |||
| 99 | |||
| 100 | |||
| 101 | =head2 MetricDefinitions => ArrayRef[L<Paws::SageMaker::MetricDefinition>] | ||
| 102 | |||
| 103 | A list of metric definition objects. Each object specifies the metric | ||
| 104 | name and regular expressions used to parse algorithm logs. Amazon | ||
| 105 | SageMaker publishes each metric to Amazon CloudWatch. | ||
| 106 | |||
| 107 | |||
| 108 | =head2 TrainingImage => Str | ||
| 109 | |||
| 110 | The registry path of the Docker image that contains the training | ||
| 111 | algorithm. For information about docker registry paths for built-in | ||
| 112 | algorithms, see Algorithms Provided by Amazon SageMaker: Common | ||
| 113 | Parameters | ||
| 114 | (https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-algo-docker-registry-paths.html). | ||
| 115 | Amazon SageMaker supports both C<registry/repository[:tag]> and | ||
| 116 | C<registry/repository[@digest]> image path formats. For more | ||
| 117 | information, see Using Your Own Algorithms with Amazon SageMaker | ||
| 118 | (https://docs.aws.amazon.com/sagemaker/latest/dg/your-algorithms.html). | ||
| 119 | |||
| 120 | |||
| 121 | =head2 B<REQUIRED> TrainingInputMode => Str | ||
| 122 | |||
| 123 | The input mode that the algorithm supports. For the input modes that | ||
| 124 | Amazon SageMaker algorithms support, see Algorithms | ||
| 125 | (https://docs.aws.amazon.com/sagemaker/latest/dg/algos.html). If an | ||
| 126 | algorithm supports the C<File> input mode, Amazon SageMaker downloads | ||
| 127 | the training data from S3 to the provisioned ML storage Volume, and | ||
| 128 | mounts the directory to docker volume for training container. If an | ||
| 129 | algorithm supports the C<Pipe> input mode, Amazon SageMaker streams | ||
| 130 | data directly from S3 to the container. | ||
| 131 | |||
| 132 | In File mode, make sure you provision ML storage volume with sufficient | ||
| 133 | capacity to accommodate the data download from S3. In addition to the | ||
| 134 | training data, the ML storage volume also stores the output model. The | ||
| 135 | algorithm container use ML storage volume to also store intermediate | ||
| 136 | information, if any. | ||
| 137 | |||
| 138 | For distributed algorithms using File mode, training data is | ||
| 139 | distributed uniformly, and your training duration is predictable if the | ||
| 140 | input data objects size is approximately same. Amazon SageMaker does | ||
| 141 | not split the files any further for model training. If the object sizes | ||
| 142 | are skewed, training won't be optimal as the data distribution is also | ||
| 143 | skewed where one host in a training cluster is overloaded, thus | ||
| 144 | becoming bottleneck in training. | ||
| 145 | |||
| 146 | |||
| 147 | |||
| 148 | =head1 SEE ALSO | ||
| 149 | |||
| 150 | This class forms part of L<Paws>, describing an object used in L<Paws::SageMaker> | ||
| 151 | |||
| 152 | =head1 BUGS and CONTRIBUTIONS | ||
| 153 | |||
| 154 | The source code is located here: L<https://github.com/pplu/aws-sdk-perl> | ||
| 155 | |||
| 156 | Please report bugs to: L<https://github.com/pplu/aws-sdk-perl/issues> | ||
| 157 | |||
| 158 | =cut | ||
| 159 |