I’m currently planning on developing software for Intel iGPUs, Nvidia, and AMD GPUs. My first question is how well does onnx work on Intel GPUs, and is it possible for both Intel and Nvidia to work with mixed precision and half precision inferencing?