AMD has launched Ryzen AI Software 1.7, enhancing user-space packages for Microsoft Windows and Linux to leverage Ryzen AI NPUs for various AI tasks, including Stable Diffusion and ONNX. This update features substantial performance improvements in its CNN/Transformer compiler, providing quicker compile times. It also introduces support for new large language models (LLMs) such as Qwen-2.5-14b-Instruct and others, along with preview support for Sparse-LLM models. Users can now benefit from long context support for hybrid execution LLM models. On the Stable Diffusion front, the update supports SD3.5-Turbo with dynamic resolution enhancements and performance improvements of up to 40% with the native BFP16 format across all compatible models. For downloads and more information on Ryzen AI Software 1.7, visit GitHub, and installation instructions for Windows and Linux can be found in the Ryzen AI documentation. This update promises improved efficiency and expanded capabilities for AI developers and enthusiasts.
Source link
