[FIXED] What does the input_shape argument to Flatten() mean?

Issue

I'm following along with the Tensorflow for Beginners tutorial:

• https://www.tensorflow.org/tutorials/quickstart/beginner

And we come to our first line of code:

Build a machine learning model

Build a tf.keras.Sequential model:

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(28, 28)),
  tf.keras.layers.Dense(128, activation='relu'),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10)
])

And i'm confused by the first line:

tf.keras.layers.Flatten(input_shape=(28, 28))

Specifically about the input_shape argument.

Why: Because i thought the Flatten operation takes the previous layer and turns it into a 1-d vector; so what is the input_shape argument mean?

So we consult the documentation

The documentation on Flatten(): https://www.tensorflow.org/api_docs/python/tf/keras/layers/Flatten

says:

Flattens the input. Does not affect the batch size.

And specifically i'm interested in the input_shape argument:

Args

data_format A string, one of channels_last (default) or channels_first. The ordering of the dimensions in the inputs. channels_last corresponds to inputs with shape (batch, ..., channels) while channels_first corresponds to inputs with shape (batch, channels, ...). When unspecified, uses image_data_format value found in your Keras config file at ~/.keras/keras.json (if exists) else 'channels_last'. Defaults to 'channels_last'.

There is no argument called input_shape.

So what does the argument mean?

Short Version

• If i were writing TF code from the documentation, how would I know that Flatten takes an input_shape argument?
• Why is the input_shape argument omitted from the documention?

In other words: there must be some basic knowledge about TensorFlow that the documentation assumes you have, but does not document itself.

Can someone help me understand this undocumented behavior, and why it is undocumented?

Research Effort

In addition to consulting the documentation, we have:

• Keras input explanation: input_shape, units, batch_size, dim, etc (isn't my question, nor answers my question)

Solution

input_shape is an optional argument that can be added to the first layer (Flatten or not) of a Keras model. Keras uses its value to create an InputLayer implicitly.

You model is equivalent to this:

model = tf.keras.models.Sequential([
  tf.keras.layers.InputLayer(input_shape=(28, 28)),
  tf.keras.layers.Flatten(),
  tf.keras.layers.Dense(128, activation='relu'),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10)
])

To quote the documentation:

When using InputLayer with the Keras Sequential model, it can be skipped by moving the input_shape parameter to the first layer after the InputLayer.

Answered By - Jafar Isbarov