These were trained on NVIDIA gpus. It is running inference on Huawei.