terminate called after throwing an instance of 'std::runtime_error' what(): The following operation failed in the TorchScript interpreter. ... RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:-2 and cuda:0! (when checking argument for argument weight in method wrapper__cudnn_convolution)
Does anyone know how to deal with this error?
I only have one GPU and made sure to send both the model and input tensors to the same device.
What does a negative device (-2) even mean?