Files
ik_llama.cpp/include
Iwan Kawrakow e345ee837d Adding q6_0_r4
We get PP-512(LLaMA-3.1-8B) = 257 t/s on a Ryzen-7950X.
2024-12-03 15:08:28 +02:00
..
2024-12-03 15:08:28 +02:00