ggerganov/llama.cpp Released Build b12255
Release b12255 of llama.cpp introduces minor fixes, specifically stopping the renaming of recurrent cache tensors in plamo2. This update ensures stability and compatibility for developers utilizing specific model architectures within the ecosystem. The release includes precompiled binaries across multiple platforms and hardware backends including CUDA, Vulkan, and macOS Metal.
## BACKGROUND
llama.cpp is an open-source library that enables efficient LLM inference in C/C++, serving as the backbone for numerous local AI tools. It is co-developed alongside GGML, a general-purpose tensor library designed for machine learning workloads.