Amazon Web Services on Wednesday added new features to its managed machine learning service Amazon SageMaker, designed to improve governance attributes within the service and adding new capabilities to its notebooks.
Notebooks in context of Amazon SageMaker are compute instances that run the Jupyter Notebook application.
Governance updates to improve granular access, improve workflow
AWS said the new features will allow enterprises to scale governance across their ML model lifecycle. As the number of machine learning models increases, it can get challenging for enterprises to manage the task of setting privilege access controls and establishing governance processes to document model information, such as input data sets, training environment information, model-use description, and risk rating.
Data engineering and machine learning teams currently use spreadsheets or ad hoc lists to navigate access policies needed for all processes involved. This can become complex as the size of machine learning teams increases within an enterprise, AWS said in a statement.
Another challenge is to monitor the deployed models for bias and ensure they are performing as expected, the company said.
To tackle these challenges, the cloud services provider has added Amazon SageMaker Role Manager to make it easier for administrators to control access and define permission for users.
With the new tool, administrators can select and edit prebuilt templates based on various user roles and responsibilities. The tool then automatically creates access policies with necessary permissions within minutes, the company said.
AWS has also added a new tool to SageMaker called Amazon SageMaker Model Cards to help data science teams shift from manual recordkeeping.
The tool provides a single location to store model information in the AWS console and it can auto-populate training details like input data sets, training environment, and training results directly into Amazon SageMaker Model Cards, the company said.
“Practitioners can also include additional information using a self-guided questionnaire to document model information (e.g., performance goals, risk rating), training and evaluation results (e.g., bias or accuracy measurements), and observations for future reference to further improve governance and support the responsible use of ML,” AWS said.
Further, the company has added Amazon SageMaker Model Dashboard to provide a central interface within SageMaker to track machine learning models.
From the dashboard, enterprise can also use built-in integrations with Amazon SageMaker Model Monitor (model and data drift monitoring capability) and Amazon SageMaker Clarify (ML bias-detection capability), the company said, adding that the end-to-end visibility will help streamline machine learning governance.
Amazon SageMaker Studio Notebook is now updated
Along with adding governance features to SageMaker, AWS has added new capabilities to Amazon SageMaker Studio Notebook to help enterprise data science teams collaborate and prepare data faster within the notebook.
A data preparation capability within Amazon SageMaker Studio Notebook will now help data science teams identify errors in data sets and correct them from inside the notebook.
The new feature allows data scientists to visually review data characteristics and remediate data quality problems, the company said, adding that the tool automatically generates charts to help users identify data-quality issues and suggests data transformations to help fix common problems.
“Once the practitioner selects a data transformation, Amazon SageMaker Studio Notebook generates the corresponding code within the notebook so it can be repeatedly applied every time the notebook is run,” the company said.
In order to make it easier for data science teams to collaborate, AWS has added a new workspace within SageMaker where data science teams can read, edit and run notebooks together in real time, the company said.
Other features to SageMaker Studio Notebook include automatic conversion of notebook code to production-ready jobs and automated validation of new machine learning models using real-time inference requests.
Additionally, AWS said that it was adding geospatial capabilities to SageMaker to allow enterprises to increase its use or role in training machine learning models.