Thread tanh logit softcapping through FlashAttention (FA2, opt-in FA3) - #3391
Open
nvegesna-netizen wants to merge 3 commits into
Open
Thread tanh logit softcapping through FlashAttention (FA2, opt-in FA3)#3391nvegesna-netizen wants to merge 3 commits into
nvegesna-netizen wants to merge 3 commits into