Skip to content

Thread tanh logit softcapping through FlashAttention (FA2 and FA3) - #3391

Merged
cyanguwa merged 42 commits into
NVIDIA:mainfrom
nvegesna-netizen:nvegesna/gemma2-softcap-core
Sep 10, 2026
Merged

Thread tanh logit softcapping through FlashAttention (FA2 and FA3)#3391
cyanguwa merged 42 commits into
NVIDIA:mainfrom
nvegesna-netizen:nvegesna/gemma2-softcap-core