mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-11-02 09:12:03 +00:00
6b64f74b55
* batched-bench : fix unified KV cache handling + pp timing * cont : run dummy token only with split KV cache