Allow NeuralNetClassifier to fit tensor labels #802 (#803)

Apply to_numpy on y in NeuralNetClassifier.check_data
before applying np.unique in case that y is a torch tensor.

Co-authored-by: Autumnii <nuaazhouyi@nuaa.edu.cn>

skorch/classifier.py

@@ -13,10 +13,9 @@
 from skorch.callbacks import EpochScoring
 from skorch.callbacks import PassthroughScoring
 from skorch.dataset import CVSplit
-from skorch.utils import get_dim
+from skorch.utils import get_dim, to_numpy
 from skorch.utils import is_dataset
 
-
 neural_net_clf_doc_start = """NeuralNet for classification tasks
 
     Use this specifically if you have a standard classification task,
@@ -117,7 +116,7 @@ def check_data(self, X, y):
             raise ValueError(msg)
         if y is not None:
             # pylint: disable=attribute-defined-outside-init
-            self.classes_inferred_ = np.unique(y)
+            self.classes_inferred_ = np.unique(to_numpy(y))
 
     # pylint: disable=arguments-differ
     def get_loss(self, y_pred, y_true, *args, **kwargs):