Torch learning notes (1) -- 19 common ways to create tensor

import torch
import numpy as np

'''  tensor （Tensor） tensor It's a multidimensional array , He is a scalar 、 vector 、 High dimensional extension of matrix ,  among , Scalar is 0 D tensor , The vector is 1 D tensor , Matrix is 2 D tensor ,RGB Image is 3 D tensor . tensor There are three kinds of creation methods ： Direct type 、 Based on numerical formula 、 Based on probability formula  '''
# 1、 Direct type 
array = np.zeros((3, 3))
tensor1 = torch.tensor(array)  # array Create as import data tensor, Do not share id
tensor2 = torch.from_numpy(array)  #  be based on array establish tensor. share id
# 2、 Based on numerical formula 
tensor3 = torch.zeros((3, 3))  #  Do not read data , Create directly 0tensor
tensor4 = torch.zeros_like(tensor3)  #  Do not read data , Create directly with tensor3 etc. size Of 0tensor
tensor5 = torch.ones((3, 3))  #  ditto 
tensor6 = torch.ones_like(tensor5)  #  ditto 
tensor7 = torch.full((3, 3), fill_value=4)  #  ditto 
tensor8 = torch.full_like(tensor7, fill_value=5)  #  ditto 
tensor9 = torch.linspace(start=1, end=10, steps=2)  #  Build equal 1 D tensor （start,end,Steps=）
tensor10 = torch.logspace(start=1, end=10, steps=3, base=2)  #  Build logarithm average 1 D tensor （start,end,Steps= Sequence length ,base= The base of a logarithmic function , Default 10）
tensor11 = torch.eye(n=3, m=4)  #  Create a two-dimensional unit diagonal matrix 
#  Based on probability formula 
tensor12 = torch.randn((3, 3))  #  Generate standard normal distribution （size= The shape of the tensor ）
tensor13 = torch.randn_like(tensor12)
tensor14 = torch.rand((3, 3))  #  Generated in [0,1) Uniform distribution of 
tensor15 = torch.rand_like(tensor14)
tensor16 = torch.randint(low=1, high=10, size=(3, 3))  #  Generated in [low,high) Uniform distribution of integers （size= The shape of the tensor ）
tensor17 = torch.randint_like(tensor16, low=1, high=10)
tensor18 = torch.randperm(n=10)  #  Generate 0 To n-1 The random arrangement of （n= Tensor length ）
tensor19 = torch.bernoulli(tensor1)  #  With input Is probability , Generate Bernoulli distribution （0-1 Distribution ,input It's a probability value ）