llama.cpp releases version b9757 with updated binaries for macOS, Linux, Android, Windows, and openEuler. The release removes unconditional softmax+sort in the top-n-sigma sampler, improving sampling efficiency. New builds support Vulkan, OpenVINO, SYCL, ROCm, and CUDA on multiple architectures, including Apple Silicon and ARM64.
llama.cpp Release b9757: New Binaries and Features
from English