should lcbench weight_decay really be on log-scale?
mb706 opened this issue · comments
the range is from 1e-5 to 1e-1, it looks to me like this should be log-transformed. or did the original lcbench-people not log-transform it?
Tracked in yahpo_gym#62
In the original paper, the data is not log-scaled... for now we will follow the original design.