Variable、Tensor、Numpy的转换

川长思鸟来 2023-10-05 08:12 24阅读 0赞

### Tensor <——> Numpy ###

Tensor与Numpy之间可以相互转换：  
代码：

**\[plain\]** [view plain][][copy][view plain]

1.  import torch
2.  import numpy as np
3.  
4.  data = np.array(\[\[1,2,3,4,5\],\[6,7,8,9,10\]\])
5.  print(data)
6.  \# numpy -> tensor
7.  data\_tensor = torch.from\_numpy(data)
8.  print(data\_tensor)
9.  \# tensor -> numpy
10. data\_numpy = data\_tensor.numpy()
11. print(data\_numpy)

输出：

**\[plain\]** [view plain][][copy][view plain]

1.  \[\[ 1 2 3 4 5\]
2.  \[ 6 7 8 9 10\]\]
3.  tensor(\[\[ 1, 2, 3, 4, 5\],
4.  \[ 6, 7, 8, 9, 10\]\], dtype=torch.int32)
5.  \[\[ 1 2 3 4 5\]
6.  \[ 6 7 8 9 10\]\]

### Variable -> Numpy ###

Variable可以直接转为Numpy：

**\[plain\]** [view plain][][copy][view plain]

1.  import torch
2.  from torch.autograd import Variable
3.  
4.  \# 定义一个Variable
5.  data = Variable(torch.Tensor(\[4\]),requires\_grad = True)
6.  \# Variable -> Numpy
7.  data\_numpy = data.detach().numpy()
8.  print('Numpy:',data\_numpy)

如果把requires\_grad改为False，得到的是Tensor：

**\[plain\]** [view plain][][copy][view plain]

1.  data\_false = Variable(data\_tensor,requires\_grad = False)
2.  print(data\_false)

输出：

`tensor([4.])`

### Numpy ->Tensor -> Variable ###

但是Numpy不可以直接转为Variable，所以要借助Tensor进行转换：

**\[plain\]** [view plain][][copy][view plain]

1.  \# Numpy --> Variable ( Numpy ->Tensor -> Variable )
2.  \# 1. Numpy ->Tensor
3.  data\_tensor = torch.from\_numpy(data\_numpy)
4.  print('Tensor:',data\_tensor)
5.  \# 2. Tensor -> Variable
6.  data\_Variable = Variable(data\_tensor,requires\_grad = True)
7.  print('Variable:',data\_Variable)

输出：

**\[plain\]** [view plain][][copy][view plain]

1.  Numpy: \[4.\]
2.  Tensor: tensor(\[4.\])Variable: tensor(\[4.\], requires\_grad=True)

若试图直接将Numpy转为Variable，

`data_Variable = Variable(data_numpy)`，会出现如下错误：  
![20190719222220363.png][]

来源:[https://www.geek-share.com/detail/2774546287.html][https_www.geek-share.com_detail_2774546287.html]

[view plain]: https://www.geek-share.com/detail/2774546287.html#
[20190719222220363.png]: https://img-blog.csdnimg.cn/20190719222220363.png
[https_www.geek-share.com_detail_2774546287.html]: https://www.geek-share.com/detail/2774546287.html