Faketensor hpu device normalization (#102512)
FakeTensor doesn't normalize device_idx and failed with below testcase.
import torch
import habana_frameworks.torch.hpu
from torch._subclasses.fake_tensor import FakeTensorMode
with FakeTensorMode.push():
a = torch.empty(1, device="hpu")
b = torch.empty(1, device="hpu:0")
result = a + b
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102512
Approved by: https://github.com/albanD