. Pytorch Basics

[1]:

1

print("hello")

stdout

hello

[2]:

1

import torch

[3]:

1

import numpy as np

[4]:

1

torch.__version__

Out[4]:

'2.10.0'

[5]:

12

arr=np.array([1,2,3,4,5])
print(arr)

stdout

[1 2 3 4 5]

[6]:

1

type(arr)

Out[6]:

numpy.ndarray

[8]:

12

# convert to pytorch array 
x=torch.from_numpy(arr)

Out[8]:

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

[9]:

12

# secondoption to convert to tensor 
x=torch.as_tensor(arr)

Out[9]:

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

[10]:

123

# lets create a 2d array
arr2d=np.arange(0.0,12)
arr2d

Out[10]:

array([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11.])

[13]:

12

# lets reshape it 
arr2d=arr2d.reshape(4,3)

[15]:

123

# convert this to  tensor
x2=torch.from_numpy(arr2d) # observe this is a 64 bit array
x2

Out[15]:

tensor([[ 0., 1., 2.], [ 3., 4., 5.], [ 6., 7., 8.], [ 9., 10., 11.]], dtype=torch.float64)

[16]:

12345

# link between numpy array and tensor 
# if we create a numpy array using below command 
arr=np.array([1,2,3,4,5])
x=torch.from_numpy(arr)
arr[0]=99

[17]:

1

print(x) # first element will change here

stdout

tensor([99, 2, 3, 4, 5])

[19]:

12345

# but if we use torch.tensor it will be a different array 
arr=np.array([1,2,3,4,5])
my_tensor=torch.tensor(arr)
my_other_tensor=torch.from_numpy(arr)
my_tensor

Out[19]:

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

[20]:

1

print(my_other_tensor)

stdout

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

[21]:

1

arr[0]=99

[22]:

1

print(my_tensor) # this will not be effected

stdout

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

[23]:

1

print(my_other_tensor) # this will be effected

stdout

tensor([99, 2, 3, 4, 5])

[ ]:

1