Hey, is it possible to use Kedro's way of handling configuration (paths, envs, etc.) for training a docker image with Sagemaker (from within a pipeline)?