Skip to content

[muon] Keep the momentum out of steps the loss scaler discards - #8435

Open
alanhuangyoo wants to merge 2 commits into
deepspeedai:masterfrom
alanhuangyoo:fix/muon-momentum-survives-overflow
Open

[muon] Keep the momentum out of steps the loss scaler discards#8435
alanhuangyoo wants to merge 2 commits into
deepspeedai:masterfrom
alanhuangyoo:fix/muon-momentum-survives-overflow

Narrow the invariant to what the guard delivers, and pin the mixed case

4ce8e62
Select commit
Loading
Failed to load commit list.