llama.cpp
c44bc1ee
- llama : keep the KV related layers on the device
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
1 year ago
llama : keep the KV related layers on the device
References
#4309 - llama : per-layer KV cache
Author
ggerganov
Parents
1fa91a48
Loading