Files
ik_llama.cpp/ggml
Iwan Kawrakow a3d1111f65 GGML_UNARY_OP_SWIGLU: Metal implementation
We get ~2% speedup for PP-512(Phi-3.5-mini).
2024-09-28 11:09:13 +03:00
..
2024-07-27 07:55:01 +02:00
2024-09-28 10:11:15 +03:00
2024-07-27 07:55:01 +02:00