Tensor_totype issues with device doesn't support double

8

Thank you, your solution is working!

>>> xpu_cap = torch.xpu.get_device_capability()
>>> xpu_cap["has_fp64"]
False

It shows that my device doesn’t support FP64.