Problem Description

When using the PARSynthesizer, SDV currently expects that the fitting and sampling are both done on machines with the same processors -- i.e. it is not possible for you to fit on a GPU-enabled device, but then sample from a CPU-only device. The feature request for this is in #1778.

Until we are able to accommodate this feature, we should at least surface a better error message for users explaining this limitation. The current error message looks like this:

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. 
If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') 
to map your storages to the CPU.

Expected behavior

Create a more generic error explaining that this feature isn't yet possible.

SamplingError: This synthesizer was created on a machine with GPU but the current machine is 
CPU-only. This feature is currently unsupported. We recommend sampling on the same GPU-enabled machine.