In sagemaker model explainability monitoring How to pass values for a shap_baseline if we have categorical valuesas features?

using this documentation I passing a single row as to shap_baseline parameter to implement explainability monitoring, a similar implementation of what is done in in this GitHub repo implementation. if I am passing a single row as input to shap_baseline parameter, the schedule is failing by concatenating 2 rows. If I ignore the shap_baseline (as it is optional), the schedule takes forever to run.