mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-08-30 09:01:18 +02:00
6b64f74b55
* batched-bench : fix unified KV cache handling + pp timing * cont : run dummy token only with split KV cache