SageMaker is a fully managed end-to-end machine learning service designed for data scientists, developers and machine learning experts. It provides a powerful framework for building, training and using host machines at scale. SageMaker integrates with other AWS services and supports a wide array of machine learning functions – along with a variety of outside vendors and data formats, including most open source tools. This allows users to build machine learning models and tools quickly. The platform also supports machine learning within the Internet of Things (IoT).
For customers already on the AWS cloud, Sagemaker is an easy choice.
Amazon SageMaker is designed to accommodate both built-in algorithms and custom training scripts. It offers a fully managed zero-setup integrated Jupyter authoring notebook instance and support for automatic model tuning, Apache Spark, along with other data modeling, machine learning and deep learning libraries and frameworks. This includes TensorFlow, MXNet, Scikit, PyTorch and Chainer.
This provides a flexible and scalable approach that supports general machine learning instances as well as GPU-powered instances. The platform also supports Docker containers and provides robust monitoring, security and API management tools. AWS offers intuitive tools that make it relatively easy for data scientists and power developers to build, manage and customize machine learning environments.
SageMaker also connects to SageMaker Ground Truth, which provides access to public and private human labelers and includes built-in workflows and interfaces for common labeling tasks. The platform handles automated data labeling and integrated workflows.
Overview and Features
Data scientists and developers with knowledge of machine learning frameworks.
Scripting Languages/Formats Supported
Most major data and file types through Jupyter Notebooks.
Connects to a wide array of AWS tools, including SageMaker Ground Truth. Supports Apache Spark, Docker containers and other tools and formats, including scikit, Pandas, NumPy, TensorFlow and Apache MXNet.
Reporting and Visualization
Offers strong visualization and reporting functions.
Amazon offers a free tier with 250 hours of t2.medium notebook usage, plus 50 hours of m3.xlarge for training, along with a combined total of 125 hours of m4.xlarge for deploying machine learning models for real-time inferencing and batch transform with Amazon SageMaker. Larger volume users pay based on data volumes and usage.
AWS Sagemaker Overview and Features at a Glance:
|Vendor and features||AWS SageMaker|
|ML Focus||Hosted platform that supports supervised and unsupervised ML and deep learning.|
|Key Features and capabilities||Zero-setup integrated Jupyter authoring notebook instance. Access to rich array of open source frameworks and tools.|
|User comments||High marks for the interface, configurability and usability. However, some say that the familiarity with AWS is crucial.|
|Pricing and licensing||Complex tiered pricing model based on ML notebook instances and cloud resources consumed.|