llama.cpp

1da7b765 - server : fix speculative decoding with context shift (#10641)

server : fix speculative decoding with context shift (#10641) * server : fix speculative decoding with context shift ggml-ci * server : take into account speculative limits ggml-ci * server : add tests