Open Models Almanac
Methodsmodel architecturetoken mixersoftmax attention

single mention · model architecture

SSMax

Uses per-head query scaling as an attention normalization adaptation.

The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See softmax attention for the methods filed beside it.

Also called Per-head query scaling with SSMax.

Filed under model architecture › token mixer › softmax attention