it will Raise RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

[xiaozhen28]

it will Raise RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! if gmm.fix(x) be used in the second time. I dont konw hao to fix it. The code as gmm = GaussianMixture(n_components, x.size(1), covariance_type="diag").cuda() gmm.fit(x) gmm.fit(x) #this will raise error

[LifelongYuan]

The error occurs if you call fit() multiple times. The improper initialization of self.var causes this problem. You could have a quick fix by adding converting it to cuda:

gmm-torch/gmm.py

Line 77 in 23eaf64

self.var = torch.nn.Parameter(torch.ones(1, self.n_components, self.n_features), requires_grad=False)