Skip to content

Cortex-M: fold constant attention-score scale into the softmax-input quantize#21027

Draft
rascani wants to merge 5 commits into
pytorch:mainfrom
rascani:cortexm-scale-fold
Draft

Cortex-M: fold constant attention-score scale into the softmax-input quantize#21027
rascani wants to merge 5 commits into
pytorch:mainfrom
rascani:cortexm-scale-fold

Commits

Commits on Jul 21, 2026