Shap DeepPlainer错误：＆＃x27; tuple＆＃x27;对象没有属性＆＃x27;设备＆＃x27;

Question

我正在尝试为我的多元时间序列模型（pytorch）构建一个解释器，如下所示：

e = shap.DeepExplainer(
        model.to('cuda:1'), 
        torch.tensor(X_test_matrix,dtype=torch.float).to('cuda:1'))

x_test_matrix是3D矩阵，因为它是时间序列数据。 但是，我收到这样的错误：

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-43-6aa8f8197e30> in <module>
      1 e = shap.DeepExplainer(
      2         model.to('cuda:1'),
----> 3         torch.tensor(X_test_matrix,dtype=torch.float).to('cuda:1'))

~/env/lib64/python3.6/site-packages/shap/explainers/_deep/__init__.py in __init__(self, model, data, session, learning_phase_flags)
     84             self.explainer = TFDeep(model, data, session, learning_phase_flags)
     85         elif framework == 'pytorch':
---> 86             self.explainer = PyTorchDeep(model, data)
     87 
     88         self.expected_value = self.explainer.expected_value

~/env/lib64/python3.6/site-packages/shap/explainers/_deep/deep_pytorch.py in __init__(self, model, data)
     55 
     56             # also get the device everything is running on
---> 57             self.device = outputs.device
     58             if outputs.shape[1] > 1:
     59                 self.multi_output = True

AttributeError: 'tuple' object has no attribute 'device'

如何解决此问题？

Answer 1

我相信问题是由于CUDA指数。

这是否为您修复？

e = shap.DeepExplainer(
        model.to(torch.device('cuda')), torch.tensor(X_test_matrix, dtype=torch.float, device=torch.device('cuda'))