SageMaker rebuilding automation

When a SageMaker Canvas dataset is automatically updated, the model does not automatically rebuild with the new data. While Canvas creates new versions of your dataset with each update, you would need to manually rebuild your model using the latest dataset version.

If you're looking to automate model retraining outside of Canvas, you can use the SageMaker SDK programmatically. You can create an Estimator object that manages the infrastructure and training task. This approach allows you to set up automated retraining workflows in your deployment environment.

For example, you could create a script that:

1. Detects when your dataset has been updated
2. Creates a SageMaker Estimator with the appropriate parameters
3. Initiates the training job with the updated dataset

While Canvas itself doesn't automatically rebuild models with fresh data, it does offer some automation capabilities. You can set up automatic batch predictions that run whenever your dataset is updated, allowing your existing model to generate predictions on new data. These automations can be managed through the ML Ops section in Canvas, where you can view, edit, or delete your automation configurations.

Note that Canvas automations only run while you're logged into the Canvas application and pause when you log out. Also, there's a limit of 20 automatic configurations allowed in Canvas.
Sources
if sagemaker dataset gets a new version automatically, does the model automatically rebuild? | AWS re:Post
Community | Fast Pre-trained Model Deployment - The code only approach
View your automations - Amazon SageMaker
Make automatic batch predictions - Amazon SageMaker AI