The following SQL programs would raise an error when predicting. It is because we do not support saving the column data which is not inside the features (i.e., petal_width in the following example).
SELECT sepal_length, sepal_width, petal_length FROM iris.train TO TRAIN sqlflow_models.OneClassSVM WITH model.kernel = "rbf" INTO iris.my_model; SELECT sepal_length, sepal_width, petal_length, petal_width FROM iris.train TO PREDICT iris.predict_result.label USING iris.my_model;