TensorFlow - How to Use Placeholders

In this article, we will learn about Placeholders in TensorFlow, how to initialize them and how to pass the data to them.

While Constants are used to feed data from inside the model, Placeholders are used to feed data from outside the model. In simpler terms, placeholders doesn't provide values while initializing, they are passed during the session.

First, Initialize two placeholders.
a = tf.placeholder(tf.float32)
b = tf.placeholder(tf.float32)
Take a look at the initialization, we aren't passing any value to the tensor. Next, let's add them.
adder_node = tf.add(a, b)
Now, Inside a session we can pass any number of inputs and the model performs operations and gives a value.
# Add two numbers
sess.run(adder_node, {a: 1.3, b: 5})
# Add two 1x2 matrices
sess.run(adder_node, {a: [2, 8], b: [3, 5]})
Below is the code that initializes tensors with placeholders and passes the values at a later stage. Play with it for better understanding.

Comments

  1. You rock guy, made in some lines easier to understand then in Udacity. Thanks

    ReplyDelete

Post a Comment

Popular posts from this blog

What is a Tensor Rank and Tensor Shape - TensorFlow

Arithmetic Operations in Tensorflow

How to Create Facebook App