mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-07-26 19:23:37 -04:00
3637576288
* server : use swa-full fo draft context ggml-ci * server : disable speculative decoding for SWA models