PyTorch101 --PyTorch基本框架

一、Tensor 张量

定义：The number of dimensions is the rank of the tensor；例：tensor([1,2,3])的秩为1

the shape of a tensor is a tuple of integers giving the size of the array along each dimension 例：tensor([1,2,3])的shape为[3]

1.构建&访问tensor

x=torch.tensor([[1,2,3],[4,5,6]])#构建tensor张量
x.dim()#返回纬度值
x.shape#返回：torch.Size([3,3]),可以当作turple使用
x[0][0]=1#赋值操作
x[0][0].item()#将PyTorch的标量转换为python标量

2.帮助构建tensor的函数

常用的函数有：

- x=torch.zeros（*size） #全为0的tensor
- x=torch.ones(*size)#全为1的tensor
- x=torch.rand(*size)#为0-1随机数
- x=torch.full（*size*， *fill_value*， ***， *out* *=* *None*， *dtype* *=* None， *layout* *=* *torch.strided*， *device* *=* *None*， *requires_grad* *=* *False*）#→ Tensor
- x=torch.from_numpy( ndarray ) #→ Tensor
- x=torch.arange( *start=0* , *end* , *step=1* , * , out=None , *dtype=None* , layout *=torch.strided* , *device=None* , *require_grad=False* )#→Tensor <u>#左闭右开区间</u>（torch.range <u>#左闭右闭</u>）
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

  

You can find a **full list of tensor creation operations** [in the documentation](https://www.google.com/url?q=https%3A%2F%2Fpytorch.org%2Fdocs%2Fstable%2Ftorch.html%23creation-ops).



#### 3.tensor的数据形式

tensor会根据数字自动设定形式，你也可以为其赋予特定格式如：int16、uint8、float32等等。

可以使用  .to( )来修改dtype，如：x0.to(torch.float32)

```python
x1=torch.zeros_like(x0) #构建一个size和dtype与x0一样的
x2=x0.new_zeros(4,5)#size自定，dtype和x0一样
x3=torch.ones(6,7).to(x0)#和上一行相同

常用的数据形式：

torch.float32: Standard floating-point type; used to store learnable parameters, network activations, etc. Nearly all arithmetic is done using this type.
torch.int64: Typically used to store indices
torch.bool: Stores boolean values: 0 is false and 1 is true
torch.float16: Used for mixed-precision arithmetic, usually on NVIDIA GPUs with tensor cores . You won’t need to worry about this datatype in this course.

4.Tensor Indexing

tensor也能进行“切片”，多维度也行