llama.cpp
0644c3be - phi-2 : scale Q instead of KQ for better precision

Commit
1 year ago
phi-2 : scale Q instead of KQ for better precision
Author
Parents
Loading