llama.cpp Update: K/V Rotation Input Guard Added

Core summary

One sentence to understand this update

llama.cpp's latest update, b9873, adds a guard for K/V rotation input, preventing issues when the buffer is unallocated and enhancing model stability.

Impact & opportunity

What this could mean

Builders can leverage this llama.cpp enhancement for more robust local LLM deployments, reducing potential runtime errors related to K/V cache management.