When users run a Teradata ML analytic function, results are stored as tables in the Teradata Database that is specified in the Teradata Vantage connection.
However, not all of these resulting tables may be persistent (in permanent storage) in the connection database. Specifically, tables that store models produced by analytic functions are non-persistent work tables (temporary tables).
The difference is that tables in permanent storage persist across different sessions, whereas temporary tables are automatically dropped at the end of a session.
Therefore, if the user establishes a Teradata Vantage connection in Python and calls an analytic function that creates an analytic model table in the Teradata Database, when the user eliminates the connection, the database session will be terminated and the model table will be automatically dropped from the database.
To preserve a non-persistent model table created by teradataml, use the copy_to function with the model as a table object input to the function, before disconnecting from the session where the model table was created.