tf.Module

TensorFlow 2 version View source on GitHub

Base neural network module class.

Inherits From: Checkpointable

A module is a named container for tf.Variables, other tf.Modules and functions which apply to user input. For example a dense layer in a neural network might be implemented as a tf.Module:

 class Dense(tf.Module):
   def __init__(self, in_features, output_features, name=None):
     super(Dense, self).__init__(name=name)
     self.w = tf.Variable(
         tf.random.normal([input_features, output_features]), name='w')
     self.b = tf.Variable(tf.zeros([output_features]), name='b')

   def __call__(self, x):
     y = tf.matmul(x, self.w) + self.b
     return tf.nn.relu(y)

You can use the Dense layer as you would expect:

d = Dense(input_features=64, output_features=10)
d(tf.ones([100, 64]))
#==> <tf.Tensor: ...>

By subclassing tf.Module instead of object any tf.Variable or tf.Module instances assigned to object properties can be collected using the variables, trainable_variables or submodules property:

d.variables
#==> (<tf.Variable 'b:0' ...>, <tf.Variable 'w:0' ...>)

Subclasses of tf.Module can also take advantage of the _flatten method which can be used to implement tracking of any other types.

All tf.Module classes have an associated tf.name_scope which can be used to group operations in TensorBoard and create hierarchies for variable names which can help with debugging. We suggest using the name scope when creating nested submodules/parameters or for forward methods whose graph you might want to inspect in TensorBoard. You can enter the name scope explicitly using with self.name_scope: or you can annotate methods (apart from __init__) with @tf.Module.with_name_scope.

class MLP(tf.Module):
  def __init__(self, input_size, sizes, name=None):
    super(MLP, self).__init__(name=name)
    self.layers = []
    with self.name_scope:
      for size in sizes:
        self.layers.append(Dense(input_size=input_size, output_size=size))
        input_size = size

  @tf.Module.with_name_scope
  def __call__(self, x):
    for layer in self.layers:
      x = layer(x)
    return x

name Returns the name of this module as passed or determined in the ctor.

name_scope Returns a tf.name_scope instance for this class.
submodules Sequence of all sub-modules.

Submodules are modules which are properties of this module, or found as properties of modules which are properties of this module (and so on).

a = tf.Module()
b = tf.Module()
c = tf.Module()
a.b = b
b.c = c
assert list(a.submodules) == [b, c]
assert list(b.submodules) == [c]
assert list(c.submodules) == []

trainable_variables Sequence of variables owned by this module and it's submodules.

variables Sequence of variables owned by this module and it's submodules.

Methods

with_name_scope

View source

Decorator to automatically enter the module name scope.

class MyModule(tf.Module):
  @tf.Module.with_name_scope
  def __call__(self, x):
    if not hasattr(self, 'w'):
      self.w = tf.Variable(tf.random.normal([x.shape[1], 64]))
    return tf.matmul(x, self.w)

Using the above module would produce tf.Variables and tf.Tensors whose names included the module name:

mod = MyModule()
mod(tf.ones([8, 32]))
# ==> <tf.Tensor: ...>
mod.w
# ==> <tf.Variable ...'my_module/w:0'>

Args
method The method to wrap.

Returns
The original method wrapped such that it enters the module's name scope.