Repository navigation

cuda: tree attention and DFlash block attention for every family, checked against float64 - #746

Closed

jschmied wants to merge 14 commits into

ashhart:mainfrom

jschmied:cuda-attn-shared

Closed

cuda: tree attention and DFlash block attention for every family, checked against float64#746
jschmied wants to merge 14 commits into
ashhart:mainfrom
jschmied:cuda-attn-shared