I just updated to pyTorch 0.4.1 and my code that was running fine on 0.4.0 is now throwing this runtime error.
<ipython-input-9-c1a4924f6dbc> in init_freq_alias_matrix(self, swapped)
167 n = n + 1
168 alias_op = sparse.coo_matrix((temp[0:n], (ii[0:n], jj[0:n])), shape=(np.prod(M), np.prod(N)))
--> 169 self.alias_op = sparse_mx_to_torch_sparse_tensor(alias_op).cuda()
170
171
RuntimeError: torch.cuda.sparse.DoubleTensor is not enabled.
How do I enable them?