tf.layers.dropout() vs tf.nn.dropout(): A Difference Introduction – TensorFlow Tutorial

In tensorflow, we can use tf.layers.dropout() and tf.nn.dropout() to create a dropout layer. In this tutorial, we will introduce their difference.

tf.layers.dropout()

You can learn how to use tf.layers.dropout() in this tutorial:

Understand TensorFlow tf.nn.dropout(): A Beginner Guide – TensorFlow Tutorial

tf.layers.dropout()

It is defined as:

__init__(
    rate=0.5,
    noise_shape=None,
    seed=None,
    name=None,
    **kwargs
)

We can find how to use rate in tf.layers.dropout()

In summation, in order to create and use dropout layer correctly, we should notice:

When training, we can set 0.5 in tf.layers.dropout() and tf.nn.dropout(). However, when testing, we should set 1.0 for tf.nn.dropout(), 0 for tf.layers.dropout().