深度学习:pytorch常见错误总结

最新推荐文章于 2025-06-19 11:12:37 发布

原创最新推荐文章于 2025-06-19 11:12:37 发布 · 2.1w 阅读

3 ·

CC 4.0 BY-SA版权

文章标签：

#cuda #深度学习

pytorch 专栏收录该内容

19 篇文章

订阅专栏

本文通过示例代码详细解析了PyTorch中CUDA张量与CPU张量不匹配导致的错误，并提供了如何避免此类错误的方法。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

1、expected CPU tensor (got CUDA tensor)

期望得到CPU类型张量，得到的却是CUDA张量类型。

很典型的错误，例如计算图中有的参数为cuda型有的参数却是cpu型就会遇到这样的错误。

>>> import torch
>>> from torch.autograd import Variable
>>> a = torch.Tensor([1])
>>> b = torch.Tensor([2])
>>> a = Variable(a)
>>> b = Variable(b)
>>> a.requires_grad = True
>>> b = b.type(torch.cuda.FloatTensor)
>>> c = a + b   # 这里a和b两个张量不在同一个空间一个在cpu中另一个在gpu中因此会引发错误
Traceback (most recent call last):
  File "C:\Users\dell\Anaconda3\envs\my-pytorch\lib\site-packages\IPython\core\interactiveshell.py", line 2862, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-16-60f555c9e9aa>", line 1, in <module>
    c = a + b
  File "C:\Users\dell\Anaconda3\envs\my-pytorch\lib\site-packages\torch\autograd\variable.py", line 813, in __add__
    return self.add(other)
  File "C:\Users\dell\Anaconda3\envs\my-pytorch\lib\site-packages\torch\autograd\variable.py", line 319, in add
    return self._add(other, False)
  File "C:\Users\dell\Anaconda3\envs\my-pytorch\lib\site-packages\torch\autograd\variable.py", line 313, in _add
    return Add.apply(self, other, inplace)
  File "C:\Users\dell\Anaconda3\envs\my-pytorch\lib\site-packages\torch\autograd\_functions\basic_ops.py", line 17, in forward
    return a.add(b)
TypeError: add received an invalid combination of arguments - got (torch.cuda.FloatTensor), but expected one of:
 * (float value)
      didn't match because some of the arguments have invalid types: (!torch.cuda.FloatTensor!)
 * (torch.FloatTensor other)
      didn't match because some of the arguments have invalid types: (!torch.cuda.FloatTensor!)
 * (torch.SparseFloatTensor other)
      didn't match because some of the arguments have invalid types: (!torch.cuda.FloatTensor!)
 * (float value, torch.FloatTensor other)
 * (float value, torch.SparseFloatTensor other)