Does ORT support multiple FPGAs side-by-side?

[ExecutionProvider]

For example, can I perform inference for my model on two or more Intel Arria 10 FPGAs?

Another related question...

I saw here ORT does not support inference on multiple GPUs:
#6216

But what about using multiple GPUs for training? Is this possible?

[ExecutionProvider]

I found the answer myself. It can be done through the OpenVINO EP.