Fix 128x softmax_lse over-allocation in the dense flash-attn-v3 path#3597
Open
Xueying-VirtueAI wants to merge 1 commit into
Open
Fix 128x softmax_lse over-allocation in the dense flash-attn-v3 path#3597Xueying-VirtueAI wants to merge 1 commit into
Xueying-VirtueAI wants to merge 1 commit into