Snapshots are an artifact of calls to
pgml.train that specify the relation_name. See training for ways to create new snapshots.
Every snapshot has an accompaning table in the
pgml schema. For example, the
id = 42 has all data recorded in the table
pgml.snaphot_42. If the test_sampling was
random for the training, the rows in the table were
ORDER BY random() when it was created so that future samples can be consistently and efficiently randomized.