Skip to main content

Creating Tensors

Summary​

  • tf.constant()
  • tf.Variable()

Creating Tensors

  1. tf.constant()
    • Once created, cannot be changed
  2. tf.Variable()
    • Can be mutated

Constant function​

Once created, the value cannot be changed

tf.constant([1, 2, 3])
# <tf.Tensor: shape=(3,), dtype=int32, numpy=array([1, 2, 3], dtype=int32)>

Variable function​

After created, the values can be changed using tf.assign or other operations

NOTE: New value should also be in the same shape and dtype

vector = tf.Variable([1, 2, 3])
# <tf.Variable 'Variable:0' shape=(3,) dtype=int32, numpy=array([1, 2, 3], dtype=int32)>

vector.assign([4, 5, 6])
vector
# <tf.Variable 'Variable:0' shape=(3,) dtype=int32, numpy=array([4, 5, 6], dtype=int32)>

vector[0].assign(1)
vector
# <tf.Variable 'Variable:0' shape=(3,) dtype=int32, numpy=array([1, 5, 6], dtype=int32)>

Variable vs Constant function​

readonly_vec = tf.constant([1, 2, 3])
vec = tf.Variable([1, 2, 3])

readonly_vec, vec

#(<tf.Tensor: shape=(3,), dtype=int32, numpy=array([1, 2, 3], dtype=int32)>,
# <tf.Variable 'Variable:0' shape=(3,) dtype=int32, numpy=array([1, 2, 3], dtype=int32)>)