Deep learning -- data operation

 torch.arange(12) Produce a result from 0 Start to 11（ Include 11） A vector of ,reshape function , Convert the vector to 3x4 Matrix

import torch
x=torch.arange(12)
x=x.reshape(3,4)
#x=torch.zeros((2,3,4))# Two 2 individual , Three elements 4 A three bit array composed of two-dimensional arrays of columns 

The first line of code below is equivalent to the above code .sum（） Function is used to sum , Below cat The function user merges the two matrices ,dim be equal to 0 When, it means to merge by line ,dim be equal to 1 When, it means merging by column

x=torch.arange(12,dtype=torch.float32).reshape(3,4)
x.sum()
y=torch.tensor([[2.0,1,4,3],[1,2,3,4],[4,3,2,1]])
torch.cat((x,y),dim=0),torch.cat((x,y),dim=1)#dim=0 Press the line / One dimensional merge ,dim=1 By column / Merge in two dimensions ,dim=2

 Here is the output ：
x=torch.arange(12,dtype=torch.float32).reshape(3,4)
x.sum()
y=torch.tensor([[2.0,1,4,3],[1,2,3,4],[4,3,2,1]])
torch.cat((x,y),dim=0),torch.cat((x,y),dim=1)#dim=0 Press the line / One dimensional merge ,dim=1 By column / Merge in two dimensions ,dim=2

Judge whether it is equal according to logical operators ：

# Judge equality by logical operators 
# x=torch.arange(0,8,dtype=torch.float32).reshape(2,2,2)
# y=torch.arange(8,16,dtype=torch.float32).reshape(2,2,2)
# torch.cat((x,y),dim=2)
y==x

 Here is the output ：
tensor([[False,  True, False,  True],
        [False, False, False, False],
        [False, False, False, False]])

Even if the shape of the matrix is different in Python It can also be operated by elements through the broadcast mechanism ：

a=torch.arange(3).reshape((3,1))
b=torch.arange(2).reshape((1,2))
a,b
a+b# Even if the shape is different , We can still perform per element operations through the broadcast mechanism , Will copy the matrix as （3,2） matrix 

 Here is the output ：
a=torch.arange(3).reshape((3,1))
b=torch.arange(2).reshape((1,2))
a,b
a+b# Even if the shape is different , We can still perform per element operations through the broadcast mechanism , Will copy the matrix as （3,2） matrix 

Between the two matrices, Cao Yong can also calculate the general data operation time ：+,-,*,**,/:
 

x=torch.tensor([1.0,2,4,8])
y=torch.tensor([2,2,2,2])
x+y,x-y,x*y,x**y

 Here is the output ：
x=torch.tensor([1.0,2,4,8])
y=torch.tensor([2,2,2,2])
x+y,x-y,x*y,x**y