Hyperparameter Optimization with GraphStorm

[dombrowsky]

I am currently developing a model to classify a large transaction graph and want to test different graph architectures. For this purpose I want perform hyperparameter optimization for the model training, to correctly determine the real predictive power in the graph structures. However, I was not able to find any information on how to perform hyperparameter optimization (with SageMaker) with any GraphStorm algorithm, besides using the GraphStorm python classes with SageMaker or using a bash script to apply grid search over the CLI.

Is there any "canonical" way or best practice on how to perfrom hyperparameter optimization with GraphStorm, preferably with Sagemaker?

Thanks for any ideas or replies.