torch.Tensor和torch.tensor之间的区别

柔情只为你懂 2021-05-12 11:49 882阅读 0赞

### 引用官网里的内容：\[“ torch.Tensor是torch.FlaotTensor ”\] ###

*  Tensor默认生成FloatTensor类型的数据，tensor默认生成LongTensor类型的数据

import torch
    
    data = [1, 2, 3, 4]
    tensor_result = torch.tensor(data)
    print("tensor：", tensor_result, "|", "tensor生成的数据类型：", tensor_result.type(), "|", "dtpye:", tensor_result.dtype)
    
    Tensor_result = torch.Tensor(data)
    print("Tensor：", Tensor_result, "|", "Tensor生成的数据类型：", Tensor_result.type(), "|", "dtpye:", Tensor_result.dtype)

**result**：  
![在这里插入图片描述][20200310161628989.png]

### `torch.Tensor 是一个类` 而 `torch.tensor是一个函数` ###

*  [torch.Tensor详情链接][torch.Tensor]
 *  Class Tensor下包含很多函数。  
    ![在这里插入图片描述][20200310161638761.png]
 *  [torch.tensor详情链接][torch.tensor]
 *  torch.tensor的函数原型：
    
        torch.tensor(data, dtype=None, device=None, requires_grad=False, pin_memory=False)
    
    使用[`torch.tensor()`][torch.tensor] 总是会从传入`参数data`中拷贝数据. 假如你有tensor数据想避免被拷贝，可以使用 [`torch.Tensor.requires_grad_()`][torch.Tensor.requires_grad] or [`torch.Tensor.detach()`][torch.Tensor.detach]. 如果你有numpy的数组，并且要避免备拷贝，可以使用[`torch.as_tensor()`][torch.as_tensor].

[20200310161628989.png]: /images/20210511/9523394d0bc942ac86baea274f1d6c75.png
[torch.Tensor]: https://pytorch.org/docs/stable/tensors.html#torch.Tensor
[20200310161638761.png]: /images/20210511/b1a194348bf6453e9a3d939a850a1e53.png
[torch.tensor]: https://pytorch.org/docs/stable/torch.html#torch.tensor
[torch.Tensor.requires_grad]: https://pytorch.org/docs/stable/tensors.html#torch.Tensor.requires_grad_
[torch.Tensor.detach]: https://pytorch.org/docs/stable/autograd.html#torch.Tensor.detach
[torch.as_tensor]: https://pytorch.org/docs/stable/torch.html#torch.as_tensor