CODEWITHSUNDEEP
tensorflow2.0
ling
ling

Reputation: 1675

Why does the Tensor's value change?

# Using Python state
x = tf.zeros([10, 10])
x += 2  # This is equivalent to x = x + 2, which does not mutate the original
        # value of x
print(x)

x changed from 0 to 2. It shows the following result of x = tf.zeros([10,10]):

<tf.Tensor: id=266, shape=(10, 10), dtype=float32, numpy=
array([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0.]], dtype=float32)>

Then it changed to the following after executing: x += 2

<tf.Tensor: id=263, shape=(10, 10), dtype=float32, numpy=
array([[2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.],
       [2., 2., 2., 2., 2., 2., 2., 2., 2., 2.]], dtype=float32)>

Why does the comment say "which does not mutate the original value of x"?

Upvotes: 1

Views: 248

Answers (1)

Dan Moldovan
Dan Moldovan

Reputation: 975

Chris Heald is right. It's eaiest to see the difference using NumPy:

import numpy as np

a = np.array(2)
b = a  # Set up an alias

a += 1  # NumPy operations are in-place - they mutate the array
print(b)  # Output: 3!

Since the ndarray's __iadd__ mutates the array in-place, any references to the array will update, so the code prints 3. NumPy arrays are more like objects in this regard.

Compare this with TF Tensors, which are immutable (code is TF 2):

import tensorflow as tf

a = tf.constant(2)
b = a  # Set up an alias

a += 1  # Tensor operations are not in-place - a new tensor is created
print(b)  # Output: 2

That prints 2, because Tensor is immutable. So they are more like primitive values.

Accessing the original value is then straightforward - just assign it to some other variable (like I did with b = a).

Another way to describe this is using lists:

l = [1]
l[0] = 2  # I can mutate the list...
l = [2]   # ... or I can create a new one

Upvotes: 2

Related Questions