I like what @LittleNyima has shared. In addition to that, you can also execute the query multiple times with LIMIT and OFFSET if the solution allows it.

def generate_dataset(cursor, limit=1000, offset=200):
    query_with_limit = query + f" LIMIT {limit} OFFSET {offset}" 
    cursor.execute(query_with_limit)
    rows = cursor.fetchall()
    if not rows:
        return tuple()
    else:
        for row in rows:
            # some transformation
            yield tuple_version_of_row