Amazon SageMaker now offers a fully managed MLflow Capability
Amazon SageMaker now offers a fully managed MLflow capability. Data scientists can use familiar MLflow constructs to organize, track, and analyze ML experiments and administrators can setup MLflow with better scalability, availability, and security.
MLflow is a popular open-source tool that helps customers manage ML experiments. Data scientists and ML engineers are already using MLflow with SageMaker. However, it required setting up, managing, and securing access to MLflow Tracking Servers. With this launch, SageMaker makes it easier for customers to set-up, and manage MLflow Tracking Servers with a couple of clicks. Customers can secure access to MLflow via AWS Identity and Access Management roles. Data scientists can use MLflow SDK to track experiments across local notebooks, IDEs, managed IDEs in SageMaker Studio, SageMaker Training Jobs, SageMaker Processing Jobs, and SageMaker Pipelines. Experimentation capabilities such as rich visualizations for run comparisons and model evaluations are available to help data scientists find the best training iteration. Models registered in MLflow automatically appear in the SageMaker Model Registry for a unified model governance experience and customers can deploy MLflow Models to SageMaker Inference without building custom MLflow containers. The integration with SageMaker allows data scientists to easily track metrics during model training ensuring reproducibility across different frameworks and environments.
Amazon SageMaker with MLflow is available in all Amazon Web Services regions where Amazon SageMaker is currently available, excluding China Regions and GovCloud (US) Regions. For pricing, see Amazon SageMaker MLflow pricing.
To get started, refer to Amazon SageMaker developer guide and blog.