Create basic tensors using tf.constant()

Let's create a scalar first:

r0_tensor = tf.constant(4)
r0_tensor

Running the cell would give you the following:

<tf.Tensor: shape=(), dtype=int32, numpy=4>

tf.constant() creates a constant tensor in the form of a tensor-like object that has:

shape - the number of elements in each of the axes of a tensor.
dtype - the data type of the elements in the tensor.
numpy - every tensor has a numpy method that lets you access the value from the tensor object.

Create a Vector

Try to create a vector now which is a 1D array passed to the tf.constant().

r1_tensor = tf.constant([3, 5.0, 6])
r1_tensor

Try it: What did you get in the output this time? Did you notice something unique?

Create a Matrix / Rank-2 Tensor

r2_tensor = tf.constant([[2,3], [4,5], [6,7]])
r2_tensor

Output:

<tf.Tensor: shape=(3, 2), dtype=int32, numpy=
array([[2, 3],
       [4, 5],
       [6, 7]], dtype=int32)>

The shape here is (3, 2) which means that we have 3 elements in the first axis(or dimension) and 2 elements in the second axis.

 

Create a Rank-3 Tensor

r3_tensor = tf.constant([
  [[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],
   [25, 26, 27, 28, 29]],])

print(r3_tensor)

Let's look at the output of these now:

Try answering the following questions after running the code above:

• What is the shape of the tensor?
• What is the data type of the elements in the tensor?
• How many axes does this tensor have?
• What is the total number of elements in the tensor?

Also, try creating a rank-4 tensor yourself, you might want to try looking at tf.zeros for an easy way out.

Use this momentum, keep going!

So, you have learned what scalars, vectors and matrices are and how you can create these using tensorflow. The next steps should be to understand the attributes and what we can do with them.

Let's dive right in!