onnxruntime
[CUDA] Fuse add bias and transpose into one kernel in Attention
#12670
Merged
Merged