jianlin.su
@Jianlin_S
Grad&Clip&EM is all you need @Kimi_Moonshot
Blog: , Cool Papers:
Interesting link:
参加 February 2025
21 フォロー中 11.6K ファン
Steepest Descent on Manifolds: 6. Muon + Double Rotation
Introduces MuonR — a Muon variant that constrains updates to left & right rotation matrices. This preserves the singular value distribution of weights, providing a clean, elegant way to maintain training stability.
もっと見る
