List your company

Masked Language Modeling on omg prot50 (val)

78Wall-clock Time (17.5 Target PPL)MUD1
75.47292.536109.6126.664Mar 18, 2026

Evaluation results

MethodLinks
MUD1Hardware=AMD MI300X, Total training steps=5,000 steps, Model Architecture=ESM-2 150M2026.037811012616.8556361.81
MUD2Hardware=AMD MI300X, Total training steps=5,000 steps, Model Architecture=ESM-2 150M2026.0378.286.2131.316.7957101.81
MuonHardware=AMD MI300X, Total training steps=5,000 steps, Model Architecture=ESM-2 150M2026.03127.8188.4236.716.849971.1
AdamWHardware=AMD MI300X, Total training steps=5,000 steps, Model Architecture=ESM-2 150M2026.03141.2245.1392.817.176466—