The output may be non-deterministic for some functions if the insertion order of the rows in the table change, or the cluster configuration (for example, different number of AMPs) changes. This is due to change in data distribution. Examples include TD_SVM, TD_GLM, TD_DecisionForest, TD_XGBoost, TD_KMeans, TD_TargetEncodingFit, and so on.
The output may be non-deterministic for some functions due to the random nature of the algorithm. This is typically controlled by the Seed argument. Examples include TD_DecisionForest, TD_KMeans, and so on.