Pythonic LLM inference on legacy GPUs using Vulkan — GPU-accelerated local AI for AMD, Intel, and NVIDIA without CUDA.
