Amazon SageMaker Experiments launches new capabilities to manage ML experiments performed in diverse IDEs
Amazon SageMaker Experiments now supports tracking and analysis of machine learning (ML) experiments performed in any IDE (e.g., SageMaker Studio, JupyterHub) or executable code (e.g., local notebooks, scripts) using the SageMaker Python SDK or Boto3. You can track the inputs, parameters, configurations, and results of your ML training iterations. You can assign, group, and organize these iterations into experiments.
Amazon SageMaker Experiments now supports tracking and analysis of machine learning (ML) experiments performed in any IDE (e.g., SageMaker Studio, JupyterHub) or executable code (e.g., local notebooks, scripts) using the SageMaker Python SDK or Boto3. You can track the inputs, parameters, configurations, and results of your ML training iterations. You can assign, group, and organize these iterations into experiments.
With SageMaker Experiments, you can start organizing, tracking, and analyzing your ML experiments from your local environment. SageMaker Experiments tracks all of the steps and artifacts involved in building a model. Tracking experiments allows you to identify effective combinations of parameters and settings for optimizing your model's performance. Additionally, you can ensure the reliability and reproducibility of developed models by recreating them from tracked experiments. This is helpful when troubleshooting production issues, or auditing your models for compliance.
SageMaker Experiments is integrated with SageMaker Studio, providing a visual interface to browse your active and past experiments, compare runs on key performance metrics, and identify the best performing models. You can create charts such as line charts, scatter plots, bar charts, and histograms to analyze your recorded experiment results. SageMaker Studio allows team members to access the same information and confirm that experiment results are consistent, making collaboration easier. You can also use SageMaker Experiments to export run visualizations and then share model evaluations with your stakeholders.
You are charged only for ingestion, retrieval, and storage of metric records. As part of the AWS Free Tier, you can get started with SageMaker Experiments for free. For more information, see Amazon SageMaker pricing page.
SageMaker Experiments is generally available in all AWS commercial regions where SageMaker Studio is available, except in China. To get started, update SageMaker Studio to the latest version and create an experiment. To learn more, visit the Sagemaker Experiments product detail page and the AWS Machine Learning Blog.