kleidiai

This release bumps the llama.cpp version to 0.2.0 and ggml to 0.21.0, incorporating numerous backend improvements, bug fixes, and new model support across various hardware accelerators like SYCL, OpenCL, Metal, and Vulkan.

This release introduces a warning mechanism for weight types lacking KleidiAI kernels. It also provides pre-compiled binaries for various platforms and hardware accelerators.

This release introduces the SME2 f32 kernel for KleidiAI, enhancing performance with dynamic scheduling. It also provides pre-compiled binaries for various platforms and hardware configurations.

This release fixes a critical bug in the KLEIDIAI backend where batched 3D inputs for MUL_MAT operations were incorrectly rejected, leading to crashes during graph scheduling. Additionally, buffer checks during weight loading were relaxed.

This release introduces KleidiAI SVE 256-bit vector-length kernel integration to enhance performance on ARM-based systems. It also provides updated binaries across a wide range of operating systems and hardware backends including CUDA, Vulkan, SYCL, and HIP.