In this ungraded lab, you will try some of the basic operations you can perform on tensors.
try:
# %tensorflow_version only exists in Colab.
%tensorflow_version 2.x
except Exception:
pass
import tensorflow as tf
import numpy as np
Lets create a single dimension numpy array on which you can perform some operation. You’ll make an array of size 25, holding values from 0 to 24.
# Create a 1D uint8 NumPy array comprising of first 25 natural numbers
x = np.arange(0, 25)
x
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,
17, 18, 19, 20, 21, 22, 23, 24])
Now that you have your 1-D array, next you’ll change that array into a tensor. After running the code block below, take a moment to inspect the information of your tensor.
# Convert NumPy array to Tensor using `tf.constant`
x = tf.constant(x)
x
<tf.Tensor: shape=(25,), dtype=int64, numpy=
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,
17, 18, 19, 20, 21, 22, 23, 24])>
As the first operation to be performed, you’ll square (element-wise) all the values in the tensor x
# Square the input tensor x
x = tf.square(x)
x
<tf.Tensor: shape=(25,), dtype=int64, numpy=
array([ 0, 1, 4, 9, 16, 25, 36, 49, 64, 81, 100, 121, 144,
169, 196, 225, 256, 289, 324, 361, 400, 441, 484, 529, 576])>
One feature of tensors is that they can be reshaped. When reshpaing, make sure you consider dimensions that will include all of the values of the tensor.
# Reshape tensor x into a 5 x 5 matrix.
x = tf.reshape(x, (5, 5))
x
<tf.Tensor: shape=(5, 5), dtype=int64, numpy=
array([[ 0, 1, 4, 9, 16],
[ 25, 36, 49, 64, 81],
[100, 121, 144, 169, 196],
[225, 256, 289, 324, 361],
[400, 441, 484, 529, 576]])>
Notice that you’ll get an error message if you choose a shape that cannot be exactly filled with the values of the given tensor.
shape to avoid an error.# Try this and look at the error
# Try to change the input to `shape` to avoid an error
tmp = tf.constant([1,2,3,4])
tf.reshape(tmp, shape=(2,2))
<tf.Tensor: shape=(2, 2), dtype=int32, numpy=
array([[1, 2],
[3, 4]], dtype=int32)>
Like reshaping, you can also change the data type of the values within the tensor. Run the cell below to change the data type from int to float
# Cast tensor x into float32. Notice the change in the dtype.
x = tf.cast(x, tf.float32)
x
<tf.Tensor: shape=(5, 5), dtype=float32, numpy=
array([[ 0., 1., 4., 9., 16.],
[ 25., 36., 49., 64., 81.],
[100., 121., 144., 169., 196.],
[225., 256., 289., 324., 361.],
[400., 441., 484., 529., 576.]], dtype=float32)>
Next, you’ll create a single value float tensor by the help of which you’ll see broadcasting in action
# Let's define a constant and see how broadcasting works in the following cell.
y = tf.constant(2, dtype=tf.float32)
y
<tf.Tensor: shape=(), dtype=float32, numpy=2.0>
Multiply the tensors x and y together, and notice how multiplication was done and its result.
# Multiply tensor `x` and `y`. `y` is multiplied to each element of x.
result = tf.multiply(x, y)
result
<tf.Tensor: shape=(5, 5), dtype=float32, numpy=
array([[ 0., 2., 8., 18., 32.],
[ 50., 72., 98., 128., 162.],
[ 200., 242., 288., 338., 392.],
[ 450., 512., 578., 648., 722.],
[ 800., 882., 968., 1058., 1152.]], dtype=float32)>
Re-Initialize y to a tensor having more values.
# Now let's define an array that matches the number of row elements in the `x` array.
y = tf.constant([1, 2, 3, 4, 5], dtype=tf.float32)
y
<tf.Tensor: shape=(5,), dtype=float32, numpy=array([1., 2., 3., 4., 5.], dtype=float32)>
# Let's see first the contents of `x` again.
x
<tf.Tensor: shape=(5, 5), dtype=float32, numpy=
array([[ 0., 1., 4., 9., 16.],
[ 25., 36., 49., 64., 81.],
[100., 121., 144., 169., 196.],
[225., 256., 289., 324., 361.],
[400., 441., 484., 529., 576.]], dtype=float32)>
Add the tensors x and y together, and notice how addition was done and its result.
# Add tensor `x` and `y`. `y` is added element wise to each row of `x`.
result = x + y
result
<tf.Tensor: shape=(5, 5), dtype=float32, numpy=
array([[ 1., 3., 7., 13., 21.],
[ 26., 38., 52., 68., 86.],
[101., 123., 147., 173., 201.],
[226., 258., 292., 328., 366.],
[401., 443., 487., 533., 581.]], dtype=float32)>
When using tf.constant(), you can pass in a 1D array (a vector) and set the shape parameter to turn this vector into a multi-dimensional array.
tf.constant([1,2,3,4], shape=(2,2))
<tf.Tensor: shape=(2, 2), dtype=int32, numpy=
array([[1, 2],
[3, 4]], dtype=int32)>
Note, however, that for tf.Variable(), the shape of the tensor is derived from the shape given by the input array. Setting shape to something other than None will not reshape a 1D array into a multi-dimensional array, and will give a ValueError.
try:
# This will produce a ValueError
tf.Variable([1,2,3,4], shape=(2,2))
except ValueError as v:
# See what the ValueError says
print(v)
The initial value's shape ((4,)) is not compatible with the explicitly supplied `shape` argument ((2, 2)).