mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-11-02 09:12:03 +00:00
* faster avx512 exp implementation * x->r * improve accuracy, handle special cases * remove `e`
741 KiB
741 KiB