fix device for input tensor (PyTorch)

Changing device for the input tensor was not added in the previous MR (!130 (merged)) which is added in this MR.

bob/bio/face/embeddings/pytorch.py

@@ -81,7 +81,7 @@ class PyTorchModel(TransformerMixin, BaseEstimator):
 
         def _transform(X):
             with torch.no_grad():
-                return self.model(X).cpu().detach().numpy()
+                return self.model(X.to(self.device)).cpu().detach().numpy()
 
         if self.memory_demanding:
             return np.array([_transform(x[None, ...]) for x in X])