On the second day of AWS re:Invent 2019, Andy Jassy (CEO, Amazon Web Services) announced half a dozen new features and tools for AWS SageMaker. It is a toolkit to help developers build, train, and deploy machine learning (ML) models quickly.
The traditional machine learning model development is a complex and iterative process. It becomes, even more, harder due to a lack of integrated tools for the entire ML workflow. But AWS came up with a handful of new updates and tools dedicated to AWS SageMaker to bring ML into the mainstream.
AWS SageMaker offers the tool for each step in the ML development process.
Amazon SageMaker Studio
Amazon SageMaker Studio is the first fully integrated development environment (IDE) for machine learning. It provides a single, web-based visual interface, which enables you to carry out all the ML development steps.
There are several new updates also, but SageMaker Studio is the pivot to others. AWS shared the new stand-out features of this unified visual interface as a special announcement. Each tool provides much more into the hands of developers to push ML models to production faster with less effort and at a much lower cost than traditional methods.
Amazon SageMaker Notebooks
If you have worked with Machine Learning models, you must have heard about Jupyter Notebooks. In general, as the name suggests, “notebooks” are the tools that allow you to create and share documents that contain live code, visualizations, and narrative text. It’s like a handbook for a particular ML model with all the necessary information.
Amazon SageMaker has launched Amazon SageMaker Notebooks; which is a one-click notebook with elastic compute serving more advanced functions such as the ability to provision compute resources without interruption. You can even transfer Notebook content to new instances.
Amazon SageMaker Experiments
ML is a highly iterative process that makes it time and money consuming. Amazon SageMaker Experiments help you organize, track, compare, and evaluate machine learning (ML) experiments and model versions.
The only goal of SageMaker Experiments is to facilitate the easiest way to create experiments, populate them with trials, and run analytics across trials and experiments.
Amazon SageMaker Model Monitor
Machine Learning is crucial when applied for real-life use cases. As models are built from large amounts of data, it’s easy to say why ML practitioners care so much about data. One might have full control over their experimental data sets; the same can’t be applicable to real-life data that the models are more likely to receive. The real-life data is an unclean set of data with a “data drift” issue; i.e., a gradual shift in the statistical nature of the data.
Amazon SageMaker Model Monitor helps you detect concept drift by monitoring models deployed to production that too automatically. You can enable the Model Monitor with a single click.
Amazon SageMaker Autopilot
Machine learning models are iterative, and this requires practitioners to run models with different variables. If you are doing this manually, you have full control over your iterated models. On the other hand, even if you follow an automated approach for the model generation, which takes care of all the heavy lifting, but provides very little visibility on how the model was created.
Amazon SageMaker Autopilot automatically trains and tunes the best machine learning models for classification or regression based on your data while allowing full control and visibility.
Amazon SageMaker Debugger
Amazon SageMaker Debugger provides full visibility into the training of machine learning models by monitoring, recording, and analyzing the tensor data that captures the state of a machine learning training job at each instance in its lifecycle.
Amazon Debugger automatically detects and alerts you against commonly encountered errors such as different gradient values getting too small or too large.