llama.cpp
b150abe8
- cuda : avoid warp_reduce for smax
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
1 year ago
cuda : avoid warp_reduce for smax
References
#5021 - ggml : add Flash Attention
Author
ggerganov
Parents
b68a1122
Loading
