tf.keras.dtensor.experimental.optimizers.AdamW

DTensor specific optimizers.

Inherits From: AdamW, Optimizer

The major changes for this class is that all the variable init logic will be mesh/layout aware.

Optimizer that implements the AdamW algorithm.

AdamW optimization is a stochastic gradient descent method that is based on adaptive estimation of first-order and second-order moments with an added method to decay weights per the techniques discussed in the paper, 'Decoupled Weight Decay Regularization' by Loshchilov, Hutter et al., 2019.

According to Kingma et al., 2014, the underying Adam method is "computationally efficient, has little memory requirement, invariant to diagonal rescaling of gradients, and is well suited for problems that are large in terms of data/parameters".

learning_rate A tf.Tensor, floating point value, a schedule that is a tf.keras.optimizers.schedules.LearningRateSchedule, or a callable that takes no arguments and returns the actual value to use. The learning rate. Defaults to 0.001.
weight_decay A tf.Tensor, floating point value. The weight decay. Defaults to 0.004.
beta_1 A float value or a constant float tensor, or a callable that takes no arguments and returns the actual value to use. The exponential decay rate for the 1st moment estimates. Defaults to 0.9.
beta_2 A float value or a constant float tensor, or a callable that takes no arguments and returns the actual value to use. The exponential decay rate for the 2nd moment estimates. Defaults to 0.999.
epsilon A small constant for numerical stability. This epsilon is "epsilon hat" in the Kingma and Ba paper (in the formula just before Section 2.1), not the epsilon in Algorithm 1 of the paper. Defaults to 1e-7.
amsgrad Boolean. Whether to apply AMSGrad variant of this algorithm from the paper "On the Convergence of Adam and beyond". Defaults to False.
name String. The name to use for momentum accumulator weights created by the optimizer.
clipnorm Float. If set, the gradient of each weight is individually clipped so that its norm is no higher than this value.
clipvalue Float. If set, the gradient of each weight is clipped to be no higher than this value.
global_clipnorm Float. If set, the gradient of all weights is clipped so that their global norm is no higher than this value.
use_ema Boolean, defaults to False. If True, exponential moving average (EMA) is applied. EMA consists of computing an exponential moving average of the weights of the model (as the weight values change after each training batch), and periodically overwriting the weights with their moving average.
ema_momentum Float, defaults to 0.99. Only used if use_ema=True. This is # noqa: E501 the momentum to use when computing the EMA of the model's weights: new_average = ema_momentum * old_average + (1 - ema_momentum) * current_variable_value.
ema_overwrite_frequency Int or None, defaults to None. Only used if use_ema=True. Every ema_overwrite_frequency steps of iterations, we overwrite the model variable by its moving average. If None, the optimizer # noqa: E501 does not overwrite model variables in the middle of training, and you need to explicitly overwrite the variables at the end of training by calling optimizer.finalize_variable_values() (which updates the model # noqa: E501 variables in-place). When using the built-in fit() training loop, this happens automatically after the last epoch, and you don't need to do anything.
jit_compile Boolean, defaults to True. If True, the optimizer will use XLA # noqa: E501 compilation. If no GPU device is found, this flag will be ignored.
**kwargs keyword arguments only used for backward compatibility.

Notes:

The default value of 1e-7 for epsilon might not be a good default in general. For example, when training an Inception network on ImageNet a current good choice is 1.0 or 0.1. Note that since Adam uses the formulation just before Section 2.1 of the Kingma and Ba paper rather than the formulation in Algorithm 1, the "epsilon" referred to here is "epsilon hat" in the paper.

The sparse implementation of this algorithm (used when the gradient is an IndexedSlices object, typically because of tf.gather or an embedding lookup in the forward pass) does apply momentum to variable slices even if they were not used in the forward pass (meaning they have a gradient equal to zero). Momentum decay (beta1) is also applied to the entire momentum accumulator. This means that the sparse behavior is equivalent to the dense behavior (in contrast to some momentum implementations which ignore momentum unless a variable slice was actually used).

iterations The number of training steps this optimizer has run.

By default, iterations would be incremented by one every time apply_gradients() is called.

learning_rate

Methods

add_variable

View source

Create an optimizer variable.

Args
shape A list of integers, a tuple of integers, or a 1-D Tensor of type int32. Defaults to scalar if unspecified.
dtype The DType of the optimizer variable to be created. Defaults to tf.keras.backend.floatx if unspecified.
initializer string or callable. Initializer instance.
name The name of the optimizer variable to be created.

Returns
An optimizer variable, in the format of tf.Variable.

add_variable_from_reference

View source

Create an optimizer variable from model variable.

Create an optimizer variable based on the information of model variable. For example, in SGD optimizer momemtum, for each model variable, a corresponding momemtum variable is created of the same shape and dtype.

Args
model_variable The corresponding model variable to the optimizer variable to be created.
variable_name The name prefix of the optimizer variable to be created. The create variables name will follow the pattern {variable_name}/{model_variable.name}, e.g., momemtum/dense_1.
initial_value The initial value of the optimizer variable, if None, the value will be default to 0.

Returns
An optimizer variable.

apply_gradients

View source

Apply gradients to variables.

Args
grads_and_vars List of (gradient, variable) pairs.

Returns
None

Raises
TypeError If grads_and_vars is malformed.

build

View source

Initialize optimizer variables.

AdamW optimizer has 3 types of variables: momentums, velocities and velocity_hat (only set when amsgrad is applied),

Args
var_list list of model variables to build AdamW variables on.

compute_gradients

View source

Compute gradients of loss on trainable variables.

Args
loss Tensor or callable. If a callable, loss should take no arguments and return the value to minimize.
var_list list or tuple of Variable objects to update to minimize loss.
tape (Optional) tf.GradientTape. If loss is provided as a Tensor, the tape that computed the loss must be provided.

Returns
A list of (gradient, variable) pairs. Variable is always present, but gradient can be None.

exclude_from_weight_decay

View source

Exclude variables from weight decays.

This method must be called before the optimizer's build method is called. You can set specific variables to exclude out, or set a list of strings as the anchor words, if any of which appear in a variable's name, then the variable is excluded.

Args
var_list A list of tf.Variables to exclude from weight decay.
var_names A list of strings. If any string in var_names appear in the model variable's name, then this model variable is excluded from weight decay. For example, var_names=['bias'] excludes all bias variables from weight decay.

finalize_variable_values

View source

Set the final value of model's trainable variables.

Sometimes there are some extra steps before ending the variable updates, such as overriding the model variables with its average value.

Args
var_list list of model variables.

from_config

View source

Creates an optimizer from its config.

This method is the reverse of get_config, capable of instantiating the same optimizer from the config dictionary.

Args
config A Python dictionary, typically the output of get_config.

Returns
An optimizer instance.

get_config

View source

Returns the config of the optimizer.

An optimizer config is a Python dictionary (serializable) containing the configuration of an optimizer. The same optimizer can be reinstantiated later (without any saved state) from this configuration.

Subclass optimizer should override this method to include other hyperparameters.

Returns
Python dictionary.

minimize

View source

Minimize loss by updating var_list.

This method simply computes gradient using tf.GradientTape and calls apply_gradients(). If you want to process the gradient before applying then call tf.GradientTape and apply_gradients() explicitly instead of using this function.

Args
loss Tensor or callable. If a callable, loss should take no arguments and return the value to minimize.
var_list list or tuple of Variable objects to update to minimize loss.
tape (Optional) tf.GradientTape.

Returns
None

update_step

View source

Update step given gradient and the associated model variable.