3 ms·
Yes, it can. It's fairly common to keep a pool of learners which aren't identical for exactly this reason - either playing against prior snapshots, forking trai
by Bartweiss 9y ago
Yes, it can. It's fairly common to keep a pool of learners which aren't identical for exactly this reason - either playing against prior snapshots, forking training, or lightly randomizing some components of the learner.