mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-07-02 09:10:21 +00:00
3637576288
* server : use swa-full fo draft context ggml-ci * server : disable speculative decoding for SWA models