Aspose.LLM for .NET runs on CPU when you expected GPU, diagnose driver, acceleration selection, binary mismatch, GpuLayers setting....Inference runs at CPU speed (5-15 tokens/sec) despite a GPU being present...present. nvidia-smi does not list the Aspose.LLM process. rocm-smi...