SageMaker Canvas provides an interface for using pre-trained models provided by SageMaker JumpStart and a feature to run jobs of SageMaker Autopilot, an AutoML feature, with no code.

Note that if you finish using SageMaker Canvas, you need to log out explicitly, or you will continue to be charged for the workspace instance.

In this article, I try to run AutoML jobs with tutorial datasets, product descriptions and shipping logs.

First, create a dataset by joining the uploaded logs and descriptions data with ID.

Next, select the target column and the columns to be used for training. The recommended model type is automatically selected, but you can change it manually.

If model is for text analysis, a single text column is a source.

When the training is finished, the comparison of predicted and actual values and the importance of each column are displayed.

You can also check the scores of other models and use them.

Trained models can be registered to model registry, deployed to real-time endpoints, and can also be shared as Autopilot models to SageMaker Studio.

If you share it, you can also check the results and deploy it from SageMaker Studio, and import a Notebook to run the workflow of Autopilot with the generated settings.