Currently the primary key of the predictions table is (model_id, entity_id, as_of_date). If we wanted to use the same model for generating predictions for a different matrix (e.g., with a different test_label_timespan), we can't write those predictions to the table.
Since we are persisting the matrix_uuid and the label_timespan in the table, might be good to add both (or just matrix_uuid) to the primary_key
Currently the primary key of the predictions table is
(model_id, entity_id, as_of_date)
. If we wanted to use the same model for generating predictions for a different matrix (e.g., with a differenttest_label_timespan
), we can't write those predictions to the table.Since we are persisting the matrix_uuid and the label_timespan in the table, might be good to add both (or just matrix_uuid) to the primary_key