RuntimeError: Error attempting to use dtype torch.float64 with layout torch.sparse_coo and device type cuda

I’m trying to load a sparse matrix on GPU,but I got this RuntimeError. There is no error when I run the same code on Windows.

import torch
import torch.sparse
import scipy.io as scio

imdb_path='./data/proMatrix_512.mat'
device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu")

mat = scio.loadmat(imdb_path)['systemMatrix']
mat = mat.tocoo()
i = torch.LongTensor([mat.row, mat.col])
v = torch.DoubleTensor(mat.data)
ASRmtx = torch.sparse.DoubleTensor(i, v, mat.shape).to(device)

Could you post an executable code snippet with randomly initialized tensors to reproduce this issue?
The adapted code from the docs seems to work on Linux and Windows for float64:

nnz = 5
dims = [5, 5, 2, 2]
I = torch.cat([torch.randint(0, dims[0], size=(nnz,)),
               torch.randint(0, dims[1], size=(nnz,))], 0).reshape(2, nnz).cuda()
V = torch.randn(nnz, dims[2], dims[3]).cuda()
size = torch.Size(dims)
S = torch.sparse.DoubleTensor(I, V.double(), size)