Make all AI models work with ROCm. Enable quantization on ROCm and run LLMs locally without expensive hardware. Make local AI accessible to everyone.