AWS offerings for MLOps
Happily, AWS provides a variety of tools to help simplify this! One nice feature is called lineage tracking. SageMaker can automatically create the lineage (3) for key artifacts, including across accounts. This includes dataset artifacts, images, algorithm specifications, data configs, training job components, endpoints, and checkpoints. This is integrated with the Experiments SDK, letting you compare experiments and results programmatically and at scale. Let’s explore this visually. We’ll even generate a visualization for you to see how all of these are connected! Check it out in the following figure.
Figure 14.5 – SageMaker automatically creates lineage tracking
As you can see, the first step in tracking your lineage is running on key SageMaker resources such as training jobs, images, and processing jobs. You can use the entities that are automatically tracked, or you can define your own entities. To generate...
