onnxruntime
787dcb7d
- Support extra addition before softmax in attention cuda kernel (#9205)
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
4 years ago
Support extra addition before softmax in attention cuda kernel (#9205) * checkin qk_add in cuda ep * enable test * added todo * review comments
References
#9205 - Support extra addition before softmax in attention cuda kernel
Author
wangyems
Parents
03276527
Loading