When building MLModel, it is set to use NPU. It seems that GPU is used during inference, but it crashes during Compile.
The stack is as follows：
Selecting any option will automatically load the page