ggerganov/llama.cpp Released Build b11925

Llama.cpp has released build b11925, introducing continuous integration improvements for ROCm release builds alongside its usual collection of cross-platform binaries. This update includes optimization for ROCm build caching using ccache dependency mode. This release continues the steady stream of infrastructure enhancements that keep llama.cpp efficient and reliable across diverse hardware accelerators. Developers relying on AMD GPUs benefit directly from streamlined build processes and better CI stability. The build incorporates pull request #31231, specifically targeting the ccache dependency mode configuration for AMD ROCm. As an incremental release, it also provides pre-compiled binaries for a wide range of architectures including CUDA, Vulkan, and Apple Silicon.