This post it’s more like a start of a recepie for things that I heavily use every day. If you are like me that have platonic love for numpy and use for years since your born then inside Pytorch you don’t have to mutch worry. Numpy it’s a awesome tool but if you need use GPU for you massive mega matrix multiplication so our beloved numpy can’t help us. For that we have Pytorch! ( You can use CuPy too if your love for numpy its very maniac ( specialy for broadcast tricks ), but in this case for a better aproach it’s best use Chainer for deep learning / Neural networks )

So, no more words, Show me the code little buddy!

Where you see Out[?] it’s because all examples was made with IPython if you don’t know what is that, trust me, go to there and get use to it. This is just a couple examples that I get it from here but off course, using Pytorch this time

import torch

3. Create a null vector of size 10 (★☆☆)¶

torch.zeros(1,10)
Out[117]: 

    0     0     0     0     0     0     0     0     0     0
[torch.FloatTensor of size 1x10]

4. How to find the memory size of any array (★☆☆)

torch.zeros(1,10).size()
Out[118]: torch.Size([1, 10])

6. Create a null vector of size 10 but the fifth value which is 1 (★☆☆)

Z = torch.zeros(1,10)
Z[0,5] = 1

Z
Out[124]: 

    0     0     0     0     0     1     0     0     0     0
[torch.FloatTensor of size 1x10]

7. Create a vector with values ranging from 10 to 49 (★☆☆)

Z = torch.arange(10,50).view(1,-1)

Z
Out[128]: 


Columns 0 to 12 
   10    11    12    13    14    15    16    17    18    19    20    21    22

Columns 13 to 25 
   23    24    25    26    27    28    29    30    31    32    33    34    35

Columns 26 to 38 
   36    37    38    39    40    41    42    43    44    45    46    47    48

Columns 39 to 39 
   49
[torch.FloatTensor of size 1x40]

8. Reverse a vector (first element becomes last) (★☆☆)¶

# Using pytorch we don't have negative indices so we need to make little
# trick using your own methods

Z = torch.arange(20,30)
Out[132]: 

   20
   21
   22
   23
   24
   25
   26
   27
   28
   29
  [torch.FloatTensor of size 10]

# we create a range with inverse indices
idx = torch.LongTensor([i for i in range(Z.size(0)-1,-1,-1)])

Out[137]: 

 9
 8
 7
 6
 5
 4
 3
 2
 1
 0
[torch.LongTensor of size 10]

# done
inverted_tensor = Z.index_select(0,idx)

inverted_tensor
Out[136]: 

 29
 28
 27
 26
 25
 24
 23
 22
 21
 20
[torch.FloatTensor of size 10]

9. Create a 3x3 matrix with values ranging from 0 to 8 (★☆☆)¶

Z = torch.arange(0,9).view(3,3)

Z
Out[139]: 

 0  1  2
 3  4  5
 6  7  8
[torch.FloatTensor of size 3x3]

10. Find indices of non-zero elements from [1,2,0,0,4,0] (★☆☆)¶

Z = torch.LongTensor([1,2,0,0,4,0])

non_zero_indices = torch.nonzero(Z)

non_zero_indices
Out[142]: 

 0
 1
 4
[torch.LongTensor of size 3x1]

11. Create a 3x3 identity matrix (★☆☆)¶ - “In×n”

Z = torch.eye(3)

Z
Out[144]: 

 1  0  0
 0  1  0
 0  0  1
[torch.FloatTensor of size 3x3]

12. Create a 3x3x3 array with random values (★☆☆)

Z = torch.randn(3,3,3)

Z
Out[146]: 

(0 ,.,.) = 
 -0.3989 -0.5808  0.2944
  0.5743  1.7362 -0.5612
 -0.2097  2.0039  0.0585

(1 ,.,.) = 
 -1.9681 -1.2579 -0.4984
  0.1440 -0.0704  1.6027
 -1.2387 -0.7036 -0.8175

(2 ,.,.) = 
  0.6146 -2.3633  1.9595
  0.6888 -0.7732  0.6254
 -0.7208 -0.4531 -0.0987
[torch.FloatTensor of size 3x3x3]

13. Create a 10x10 array with random values and find the minimum and maximum values (★☆☆)

Z = torch.randn(10,10)

Zmin, Zmax = Z.min(), Z.max()

Zmin, Zmax
Out[149]: (-1.7579905986785889, 2.0110836029052734)