2 ms·Or train on the GPU and run inference on the CPU instances.by nl 11y agoOr train on the GPU and run inference on the CPU instances.