Stephen Roller
@stephenroller
Joined February 2008
1.3K Following 5.7K Followers
Some teams use sweeps, heuristics, or scaling laws to determine their training LR. At Character, we just have Noam Shazeer dial it to the right value.
