← Back to Architecture

logit softcap

Architecture
Used in
6 PRs
Best BPB
0.9076
Avg BPB
1.0970

Hyperparameters Across PRs

pr_numberparameters
211{"softcap":30}
367{"value":30}
828{"value":30}
970{"degree":5,"cap":30}
1263{"softcap":30}
1749{"value":30}