[Relax][cuDNN] Do not offload causal / non-fp16 attention, and fix the default softmax scale - #20078
Open
YangXu1990uiuc wants to merge 1 commit into
Open
[Relax][cuDNN] Do not offload causal / non-fp16 attention, and fix the default softmax scale#20078YangXu1990uiuc wants to merge 1 commit into
YangXu1990uiuc wants to merge 1 commit into
