please use torch.load with map_location='cpu' to map your storages to the CPU. code example
Example: RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available()
model = torch.load('mymodel.pt',map_location=torch.device('cpu'))