[CuTe, Fwd] Fix forward compile key churn when max_seqlen is a tensor - #2762

[CuTe, Fwd] Fix forward compile key churn when max_seqlen is a tensor#2762

jayhshah merged 1 commit into

Dao-AILab:mainDao-AILab/flash-attention:mainfrom

eamonn-zh:fix-fa4-forward-tensor-max-seqlen-cacheeamonn-zh/flash-attention:fix-fa4-forward-tensor-max-seqlen-cacheCopy head branch name to clipboard