I'm running sagemaker-spark(1.0.4 shaded inside my jar) on EMR(5.12.0) cluster and I encountered an error when trying to start XGBoost training job as in this notebook example:
ClientError: Parameter seed expects integer input.
Simply changing the seed parameter in XGBoostSageMakerEstimator class to Int works and the job finishes.
EDIT: Forgot to mention I'm actually setting the parameter to 0 (default value).