pytorch---关于tensor的一些常规知识点

1、创建一个服从均匀分布或者随机分布的张量，其大小可变，可一维，二维

tensor.rand（均匀） ：区间（0，1）
与tensor.randn（正太分布）：标准差为0，方差为1
参数：
size (int…) – a sequence of integers defining the shape of the output tensor. Can be a variable number of arguments or a collection like a list or tuple.

y = torch.rand(4)
y1 = torch.rand(2,3)
print(y)
print(y1)

x = torch.randn(4)
x1 = torch.randn(2,3)
print(x)
print(x1)

输出结果：

tensor([0.4162, 0.4271, 0.0685, 0.1288])
tensor([[0.9016, 0.5871, 0.0199],
        [0.4834, 0.1247, 0.3603]])
        
tensor([ 0.8691,  1.4418,  1.5524, -0.8321])
tensor([[-1.2848,  0.1464,  0.6162],
        [-0.2246,  1.0506, -0.2157]])

2、torch.range:返回一个一维张量，左开右闭，可带步长
torch.arange:返回一个一维张量，左开右闭，可带步长
参数：
start (Number) – the starting value for the set of points. Default: 0.
end (Number) – the ending value for the set of points
step (Number) – the gap between each pair of adjacent points. Default: 1.

z = torch.arange(1,6)
print(z)

结果如下：

tensor([1, 2, 3, 4, 5])

小技巧：
返回一维张量的有：linspace,logspace,arange,range
形状可变的有：ones，zeros，rand, randn

二、numpy 与tensor之间的转换

1. 由Numpy array数组转为张量

import numpy as np
import torch

numpy_data = np.array([1,4,5])
numpy_to_tensor = torch.from_numpy(numpy_data)

print(numpy_to_tensor)

结果如下：

2、 由张量变换为Numpy array数组

tensor_data = torch.randn(2, 4)
tensor_to_numpy = tensor_data.numpy()
print(tensor_data)
print(tensor_to_numpy)