[shchur]

Summary

After the move from the SageMaker SDK to boto3, the logs no longer named the resources AG-Cloud creates. Users had to look in the console to find their endpoint or transform job.

• Resource creation. Creating SageMaker training job / endpoint / batch transform job \{name}`. The endpoint line also shows instance_type=...orserverless`.
• Completion (with wait=True). Training job ... completed, Endpoint ... is in service, Batch transform job ... completed.
• Outputs after training. Model artifacts saved to ... is logged only when the predictor is kept, so it is skipped for FoundationModel.predict, where the tarball is empty. Predictions saved to {predictions_path} is logged for fit_predict / FM predict. Before this, it only appeared in the container logs.
• Uploads. Uploaded {train_data, tuning_data, images, test data, predictor} to s3://... replaces the "Uploading..." / "uploaded successfully" pairs. Internal uploads (ag_args, serving code, training code) are not logged.
• Removed the "Creating inference model..." messages. The model name always matches the endpoint or job name.
• Docs. The tabular and timeseries tutorials now quote the new training job message.

Testing

Lint only; CI runs the tests.

[github-actions]

Job PR-296-58a4300 is done.
Docs are uploaded to https://d12sc05jpx1wj5.cloudfront.net/PR-296/58a4300/index.html